RecSys-public.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Requirement already satisfied: implicit in /mnt/workspace/.local/lib/python3.5/site-packages (0.3.8)\n",
      "Requirement already satisfied: numpy in /usr/local/lib/python3.5/dist-packages (from implicit) (1.14.5)\n",
      "Requirement already satisfied: tqdm in /mnt/workspace/.local/lib/python3.5/site-packages (from implicit) (4.31.1)\n",
      "Requirement already satisfied: scipy>=0.16 in /usr/local/lib/python3.5/dist-packages (from implicit) (1.1.0)\n",
      "\u001b[33mYou are using pip version 18.0, however version 19.0.3 is available.\n",
      "You should consider upgrading via the 'pip install --upgrade pip' command.\u001b[0m\n"
     ]
    }
   ],
   "source": [
    "# Install the library we need for the notebook\n",
    "!pip3 install --user 'implicit'"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "# Building a music recommender system (Spring 2019)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "As its name implies, a recommender system is a tool that helps predicting what a user may or may not like among a list of given items. In some sense, you can view this as an alternative to content search, as recommendation engines help users discover products or content that they may not come across otherwise. For example, Facebook suggests friends and pages to users. Youtube recommends videos which users may be interested in. Amazon suggests the products which users may need... Recommendation engines engage users to services, can be seen as a revenue optimization process, and in general help maintaining interest in a service.\n",
    "\n",
    "In this notebook, we study how to build a simple recommender system: we focus on music recommendations, and we use a simple algorithm to predict which items users might like, that is called ALS, alternating least squares.\n",
    "\n",
    "## Goals\n",
    "\n",
    "In this lecture, we expect students to:\n",
    "\n",
    "- Revisit (or learn) recommender algorithms\n",
    "\n",
    "- Understand the idea of Matrix Factorization and the ALS algorithm\n",
    "\n",
    "- Build a simple model for a real usecase: music recommender system\n",
    "\n",
    "- Understand how to validate the results\n",
    "\n",
    "## Steps\n",
    "\n",
    "We assume students to work outside lab hours on the learning material. These are the steps by which we guide students, during labs, to build a good basis for the end-to-end development of a recommender system:\n",
    "\n",
    "* Inspect the data using Pandas, and build some basic, but very valuable knowledge about the information we have at hand\n",
    "* Formally define what is a sensible algorithm to achieve our goal: given the \"history\" of user taste for music, recommend new music to discover. Essentialy, we want to build a statistical model of user preferences such that we can use it to \"predict\" which additional music the user could like\n",
    "* With our formal definition at hand, we will learn different ways to implement such an algorithm. Our goal here is to illustrate what are the difficulties to overcome when implementing an algorithm like Matrix Factorization\n",
    "* Finally, we will focus on an existing implementation, available in ```Implicit```, which we will use out of the box to build a reliable statistical model\n",
    "\n",
    "Now, you may think at this point we will be done!\n",
    "\n",
    "Well, you'd better think twice: one important topic we will cover in all our Notebooks is **how to validate the results we obtain**, and **how to choose good parameters to train models** especially when using an \"opaque\" library for doing the job. As a consequence, we will focus on the statistical validation of our recommender system.\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "# 1. Data\n",
    "\n",
    "Understanding data is one of the most important part when designing any machine learning algorithm. In this notebook, we will use a data set published by Audioscrobbler - a music recommendation system for last.fm. Audioscrobbler is also one of the first internet streaming radio sites, founded in 2002. It provided an open API for scrobbling, or recording listeners' plays of artists' songs. last.fm used this information to build a powerful music recommender engine.\n",
    "\n",
    "## 1.1. Data schema\n",
    "\n",
    "Unlike a rating dataset which contains information about users' preference for products (one star, 3 stars, and so on), the datasets from Audioscrobbler only has information about events: specifically, it keeps track of how many times a user played songs of a given artist and the names of artists. That means it carries less information than a rating: in the literature, this is called explicit vs. implicit ratings.\n",
    "\n",
    "### Reading material\n",
    "\n",
    "- [Implicit Feedback for Inferring User Preference: A Bibliography](http://people.csail.mit.edu/teevan/work/publications/papers/sigir-forum03.pdf)\n",
    "- [Comparing explicit and implicit feedback techniques for web retrieval: TREC-10 interactive track report](http://trec.nist.gov/pubs/trec10/papers/glasgow.pdf)\n",
    "- [Probabilistic Models for Data Combination in Recommender Systems](http://mlg.eng.cam.ac.uk/pub/pdf/WilGha08.pdf)\n",
    "\n",
    "The data we use in this Notebook is available in 3 files (these files are stored in our HDFS layer, in the directory  ```/datasets/lastfm```):\n",
    "\n",
    "- **`user_artist_data.txt`**: It contains about 140,000+ unique users, and 1.6 million unique artists. About 24.2 million users' plays of artists' are recorded, along with their count. It has 3 columns separated by spaces: \n",
    "\n",
    "| UserID | ArtistID | PlayCount |\n",
    "|----|----|----|\n",
    "| ...|...|...|\n",
    "\n",
    "\n",
    "- **`artist_data.txt`** : It prodives the names of each artist by their IDs. It has 2 columns separated by tab characters (`\\t`).\n",
    "\n",
    "| ArtistID | Name |\n",
    "|---|---|\n",
    "|...|...|\n",
    "\n",
    "- **`artist_alias.txt`**: Note that when plays are scrobbled, the client application submits the name of the artist being played. This name could be misspelled or nonstandard. For example, \"The Smiths\", \"Smiths, The\", and \"the smiths\" may appear as distinct artist IDs in the data set, even though they are plainly the same. `artist_alias.txt` maps artist IDs that are known misspellings or variants to the canonical ID of that artist. The data in this file has 2 columns separated by tab characters (`\\t`).\n",
    "\n",
    "| MisspelledArtistID | StandardArtistID |\n",
    "|---|---|\n",
    "|...|...|\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "## 1.2. Understanding data: simple descriptive statistic\n",
    "\n",
    "In order to choose or design a suitable algorithm for achieving our goals, given the data we have, we should first understand data characteristics. To start, we import the necessary packages to work with regular expressions, Data Frames, and other nice features of our programming environment."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [],
   "source": [
    "%matplotlib inline\n",
    "import os\n",
    "import sys\n",
    "import re\n",
    "import random\n",
    "import matplotlib\n",
    "import implicit\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "\n",
    "from time import time\n",
    "\n",
    "os.environ[\"OPENBLAS_NUM_THREADS\"] = \"1\"   # Required by implicit\n",
    "\n",
    "base = \"/mnt/datasets/lastfm/\""
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### Question 1\n",
    "\n",
    "#### Question 1.0 (Non-grading)\n",
    "\n",
    "Using Pandas, load data from `/datasets/lastfm/user_artist_data.txt` (407 MB) and show the first 20 entries (via function `show()`).\n",
    "\n",
    "```bash \n",
    "$ du -sh /mnt/datasets/lastfm/user_artist_data.txt\n",
    "407M     /mnt/datasets/lastfm/user_artist_data.txt\n",
    "```\n",
    "\n",
    "For this Notebook, from a programming point of view, we are given the schema for the data we use, which is as follows:\n",
    "\n",
    "```\n",
    "userID: long int\n",
    "artistID: long int\n",
    "playCount: int\n",
    "```\n",
    "\n",
    "Each line of the dataset contains the above three fields, separated by a \"white space\".\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>userID</th>\n",
       "      <th>artistID</th>\n",
       "      <th>playCount</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1000002</td>\n",
       "      <td>1</td>\n",
       "      <td>55</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1000002</td>\n",
       "      <td>1000006</td>\n",
       "      <td>33</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1000002</td>\n",
       "      <td>1000007</td>\n",
       "      <td>8</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1000002</td>\n",
       "      <td>1000009</td>\n",
       "      <td>144</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1000002</td>\n",
       "      <td>1000010</td>\n",
       "      <td>314</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>1000002</td>\n",
       "      <td>1000013</td>\n",
       "      <td>8</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>1000002</td>\n",
       "      <td>1000014</td>\n",
       "      <td>42</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>1000002</td>\n",
       "      <td>1000017</td>\n",
       "      <td>69</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>1000002</td>\n",
       "      <td>1000024</td>\n",
       "      <td>329</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>1000002</td>\n",
       "      <td>1000025</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>1000002</td>\n",
       "      <td>1000028</td>\n",
       "      <td>17</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>1000002</td>\n",
       "      <td>1000031</td>\n",
       "      <td>47</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>1000002</td>\n",
       "      <td>1000033</td>\n",
       "      <td>15</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>1000002</td>\n",
       "      <td>1000042</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>1000002</td>\n",
       "      <td>1000045</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>1000002</td>\n",
       "      <td>1000054</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>1000002</td>\n",
       "      <td>1000055</td>\n",
       "      <td>25</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>1000002</td>\n",
       "      <td>1000056</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>1000002</td>\n",
       "      <td>1000059</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>1000002</td>\n",
       "      <td>1000062</td>\n",
       "      <td>71</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     userID  artistID  playCount\n",
       "0   1000002         1         55\n",
       "1   1000002   1000006         33\n",
       "2   1000002   1000007          8\n",
       "3   1000002   1000009        144\n",
       "4   1000002   1000010        314\n",
       "5   1000002   1000013          8\n",
       "6   1000002   1000014         42\n",
       "7   1000002   1000017         69\n",
       "8   1000002   1000024        329\n",
       "9   1000002   1000025          1\n",
       "10  1000002   1000028         17\n",
       "11  1000002   1000031         47\n",
       "12  1000002   1000033         15\n",
       "13  1000002   1000042          1\n",
       "14  1000002   1000045          1\n",
       "15  1000002   1000054          2\n",
       "16  1000002   1000055         25\n",
       "17  1000002   1000056          4\n",
       "18  1000002   1000059          2\n",
       "19  1000002   1000062         71"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Read the .csv file (might take a while)\n",
    "userArtistDF = pd.read_csv(base + 'user_artist_data.txt', \n",
    "                           sep=' ', \n",
    "                           header=None, \n",
    "                           names=['userID', 'artistID', 'playCount'])\n",
    "\n",
    "# Cast each column to its correct datatype\n",
    "userArtistDF['userID'] = userArtistDF['userID'].astype(np.int64, errors='ignore')\n",
    "userArtistDF['artistID'] = userArtistDF['artistID'].astype(np.int64, errors='ignore')\n",
    "userArtistDF['playCount'] = userArtistDF['playCount'].astype(np.int32, errors='ignore')\n",
    "\n",
    "# Display the first 20 records\n",
    "userArtistDF[:20]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "#### Question 1.1: \n",
    "<div class=\"alert alert-info\">\n",
    "How many distinct users do we have in our data? \n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Total n. of users:  148111\n"
     ]
    }
   ],
   "source": [
    "uniqueUsers = userArtistDF['userID'].nunique()\n",
    "print(\"Total n. of users: \", uniqueUsers)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "#### Question 1.2\n",
    "<div class=\"alert alert-info\">\n",
    "How many distinct artists do we have in our data ?\n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Total n. of artists:  1631028\n"
     ]
    }
   ],
   "source": [
    "uniqueArtists = userArtistDF['artistID'].nunique()\n",
    "print(\"Total n. of artists: \", uniqueArtists)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "-"
    }
   },
   "source": [
    "We just discovered that we have a total of 148,111 users in our dataset. Similarly, we have a total of 1,631,028 artists in our dataset. \n",
    "\n",
    "One thing we can see here is that Pandas provides very concise and powerful methods for data analytics."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Question 1.3 (Extra/Optional)\n",
    "<div class=\"alert alert-info\">\n",
    "What are the maximum and minimum play counts ?\n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The user with the most plays has reached 674412  plays.\n",
      "The user with the least plays has reached 1  play.\n"
     ]
    }
   ],
   "source": [
    "# Group by user ID's and aggregate over the sum of play counts\n",
    "userActivity = userArtistDF.groupby([\"userID\"], as_index = False)[\"playCount\"].sum()\n",
    "\n",
    "print(\"The user with the most plays has reached\",userActivity[\"playCount\"].max(), \" plays.\")\n",
    "print(\"The user with the least plays has reached\",userActivity[\"playCount\"].min(), \" play.\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "-"
    }
   },
   "source": [
    "Next, we might want to understand better user activity and artist popularity.\n",
    "\n",
    "Here is a list of simple descriptive queries that helps us reaching these purposes:\n",
    "\n",
    "* How many times each user has played a song? This is a good indicator of who are the most active users of our service. Note that a very active user with many play counts does not necessarily mean that the user is also \"curious\"! Indeed, she could have played the same song several times.\n",
    "* How many play counts for each artist? This is a good indicator of the artist popularity. Since we do not have time information associated to our data, we can only build a, e.g., top-10 ranking of the most popular artists in the dataset. Later in the notebook, we will learn that our dataset has a very \"loose\" definition about artists: very often artist IDs point to song titles as well. This means we have to be careful when establishing popular artists. Indeed, artists whose data is \"well formed\" will have the correct number of play counts associated to them. Instead, artists that appear mixed with song titles may see their play counts \"diluted\" across their songs."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### Question 2\n",
    "\n",
    "#### Question 2.1\n",
    "<div class=\"alert alert-info\">\n",
    "How many times each user has played a song? Show 5 samples of the result.\n",
    "</div>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<div class=\"alert alert-success\" style=\"background-color: AliceBlue; border-color: CornflowerBlue \"> \n",
    "The needed dataframe already obtained in <b>Question 1.3</b>\n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>userID</th>\n",
       "      <th>playCount</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>90</td>\n",
       "      <td>537</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>98</td>\n",
       "      <td>186</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>116</td>\n",
       "      <td>20</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>120</td>\n",
       "      <td>65</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>128</td>\n",
       "      <td>133</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   userID  playCount\n",
       "0      90        537\n",
       "1      98        186\n",
       "2     116         20\n",
       "3     120         65\n",
       "4     128        133"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Compute user activity\n",
    "# We are interested in how many playcounts each user has scored.\n",
    "# Group the dataframe according to userID and then aggregate by summing over the playCount\n",
    "\n",
    "# Already obtained in 1.3\n",
    "userActivity[0:5]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Visualizing User Play Count Distribution"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.collections.PathCollection at 0x7fd593354828>"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x432 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots()\n",
    "fig.set_size_inches(12, 6)\n",
    "\n",
    "ax.set_xlabel(\"User ID\")\n",
    "ax.set_ylabel(\"Play Count\")\n",
    "ax.grid(True)\n",
    "ax.set_title(\"The Total Number of Play for Every User ID\", size = 15)\n",
    "ax.scatter(userActivity['userID'],userActivity['playCount'], cmap = 'magma', c = userActivity['playCount'], s = 7)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<div class=\"alert alert-success\" style=\"background-color: AliceBlue; border-color: CornflowerBlue \"> \n",
    "When we visualize the play counts for all the user ID's, we can see that the ID's are separated where there are mostly unused ID's and we can also observe that the vast majority of the play counts are below 100K with some cases above that. However, we can observe this phenomenon in a better way by constructing the <b>Comulative distribution function</b> in the next question.\n",
    "</div>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "#### Question 2.2\n",
    "<div class=\"alert alert-info\">\n",
    "Plot CDF (or ECDF) of the number of play counts per User ID.  \n",
    "\n",
    "Explain and comment the figure you just created:   \n",
    "<ul>\n",
    "<li>for example, look at important percentiles (10%, 25%, median, 75%, tails such as >90%) and cross check with what you have found above to figure out if the result is plausible. </li>\n",
    "<li>discuss about your users, with respect to the application domain we target in the notebook: you will notice that for some users, there is very little interaction with the system, which means that maybe reccommending something to them is going to be more difficult than for other users who interact more with the system. </li>\n",
    "<li>look at outliers and reason about their impact on your reccommender algorithm</li>\n",
    "</ul>\n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x432 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots()\n",
    "fig.set_size_inches(12, 6)\n",
    "\n",
    "# Plot the cumulative distribution\n",
    "userActivity[\"playCount\"].hist(cumulative=True, density=1, bins='auto',ax=ax)\n",
    "\n",
    "colors = {0:\"red\", 1: \"orange\", 2:\"yellow\", 3:\"darkgreen\", 4:\"cyan\"}\n",
    "\n",
    "# Calculate and plot the 5 important percentiles\n",
    "for i, quantile in enumerate([.1, 0.25, .5, .75, .9]):\n",
    "    ax.axvline(userActivity[\"playCount\"].quantile(quantile), label='%.0fth perc.' % (quantile*100), color = colors[i])\n",
    "    ax.axhline(quantile, linestyle='-')\n",
    "    \n",
    "ax.semilogx()\n",
    "ax.set_title('ECDF of number of play counts per User ID')\n",
    "ax.set_xlabel('Play Counts')\n",
    "ax.set_ylabel('ECDF')\n",
    "ax.legend()\n",
    "ax.grid(False)\n",
    "ax.annotate('MEDIAN', (1,0.5), size = 15)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<div class=\"alert alert-success\" style=\"background-color: AliceBlue; border-color: CornflowerBlue \"> \n",
    "The obtained plot shows a <b>commulative distribution graph</b> for the <b>play count per user</b>. We hereby focus on <b>5</b> different percentiles (10%, 25%, 50%, 75%, 90%) and we can extract the following observations:<br>\n",
    "* 10% of userID's have played music about 34 times or less.<br>\n",
    "* 25% of userID's have played music about 204 times or less.<br>\n",
    "* 50% of userID's have played music about 892 times or less (<b>Median</b>).<br>\n",
    "* 75% of userID's have played music about 2800 times or less.<br>\n",
    "* 90% of userID's have played music about 6484 times or less.<br>\n",
    "This shows us that generally there is a strong overall interaction with the system as half the users actually have played music <b>around a thousand</b> times. This means that recommending music for this majority of the users should be just fine because they have provided a rich <b>implicit feedback</b> to the system. On the contrary, some users do not have this property where 10% of the total population just above 30 songs or less. The latter will make it more difficult to recommend music for them because we haven't collected a large amount of implicit feedback.<br> The users that have a high play count may reach 674412 plays. On the other hand, there are also users that have only one play count value. Such kind of outliers (<b>having very low counts</b>) can either be bots or just one-time users and cannot benifit from the potential of the recommender system.\n",
    "</div>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "#### Question 2.3\n",
    "\n",
    "<div class=\"alert alert-info\">\n",
    "How many play counts for each artist? Plot CDF or ECDF of the result.  \n",
    "\n",
    "Similarly to the previous question, you need to comment and interpret your result: what is the figure telling you?\n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Compute artist popularity\n",
    "# We are interested in how many playcounts per artist\n",
    "# ATTENTION! Grouping by artistID may be problematic, as stated above.\n",
    "\n",
    "# First we obtain a dataframe that groups the artist and aggregates over the total play count\n",
    "artistPopularity = userArtistDF.groupby(['artistID'], as_index = False)['playCount'].sum()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "scrolled": false,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x432 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots()\n",
    "fig.set_size_inches(12, 6)\n",
    "# Plot the cumulative distribution (100000 bins)\n",
    "artistPopularity[\"playCount\"].hist(cumulative=True, density=1, bins=100000, ax=ax)\n",
    "\n",
    "colors = {0:\"red\", 1: \"orange\", 2:\"yellow\", 3:\"darkgreen\", 4:\"cyan\"}\n",
    "\n",
    "# Calculate and plot the 5 important percentiles\n",
    "for i, quantile in enumerate([.1, 0.25, .5, .75, .9]):\n",
    "    ax.axvline(artistPopularity[\"playCount\"].quantile(quantile), label='%.0fth perc.' % (quantile*100), color = colors[i])\n",
    "    ax.axhline(quantile, linestyle='-')\n",
    "\n",
    "    \n",
    "ax.semilogx()\n",
    "ax.set_title('ECDF of number of play counts per Artist ID')\n",
    "ax.set_xlabel('Play Counts')\n",
    "ax.set_ylabel('ECDF')\n",
    "ax.grid(False)\n",
    "ax.annotate('MEDIAN', (1,0.5), size = 15)\n",
    "ax.legend(loc = 'lower right')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<div class=\"alert alert-success\" style=\"background-color: AliceBlue; border-color: CornflowerBlue \"> \n",
    "The plot above shows the <b>commulative distribution function</b> of the <b>play counts per artist ID</b>. This plot has a different shape from the one studying the user activity. Before going to the analysis of the result, lets extract the numerical data:<br>\n",
    "* 10% of artistID's have played music about 1 time or less.<br>\n",
    "* 25% of artistID's have played music about 1 time or less.<br>\n",
    "* 50% of artistID's have played music about 3 times or less (<b>Median</b>).<br>\n",
    "* 75% of artistID's have played music about 11 times or less.<br>\n",
    "* 90% of artistID's have played music about 45 times or less.<br>\n",
    "We can easily conclude that the data is <b>skewed</b> towards the lower playcount side which actually draws suspicion. And this begs the question: <b>\"can 90% of all artist actually have less than 45 plays?!\"</b> Well this is highly unlikely. But as mentioned before in <b>Question 1</b> regarding the fact that serveral artistID's may correspond to the same artist (artistID's may correspond to a song or a typo in the name of the artist) is emphasized in the plot obtained, since we have a huge portion of artistID (more than 90%) having very low play records. Thus, these artistID are most likely to represent typos and songs and hence the data we currently have is heavily diluted and hides the actual characteristics and statistics of the performance and popularity of artists on the system.<br>\n",
    "<b>NOTE</b>: it is also important to point out that we have more than <b>1.6 million artist IDs</b> where only <b>163 K</b> have more than 45 play record.\n",
    "</div>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "#### Question 2.4\n",
    "<div class=\"alert alert-info\">\n",
    "Plot a bar chart to show top 10 artists In terms of absolute play counts.  \n",
    "\n",
    "Comment the figure you just obtained: \n",
    "<ul>\n",
    "<li>are these reasonable results?</li>\n",
    "<li>is looking at top-10 artists enough to learn more about your data?</li>\n",
    "<li>do you see anything strange in the data?</li>\n",
    "</ul>\n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x432 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x432 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "topN = 10\n",
    "\n",
    "# Sort the artistID's according to decreasing playCount\n",
    "rankedArtists = artistPopularity.sort_values(by = 'playCount', ascending=False)\n",
    "\n",
    "# Extract the top 10 artists\n",
    "topNArtists = rankedArtists[:topN]\n",
    "\n",
    "# Extract the bottom 10 artists\n",
    "bottomNArtists = rankedArtists[-topN:]\n",
    "\n",
    "\n",
    "fig, ax = plt.subplots()\n",
    "\n",
    "fig.set_size_inches(12,6)\n",
    "\n",
    "topNArtists.set_index(\"artistID\")[\"playCount\"].plot(kind = 'barh', x='playCount', y='artistID', ax=ax)\n",
    "\n",
    "ax.set_xlabel('Play Count')\n",
    "ax.set_ylabel('Artist')\n",
    "ax.set_title('Top-10 Artist ID per play counts')\n",
    "\n",
    "fig2, ax2 = plt.subplots()\n",
    "\n",
    "fig2.set_size_inches(12,6)\n",
    "\n",
    "bottomNArtists.set_index(\"artistID\")[\"playCount\"].plot(kind = 'barh', x='playCount', y='artistID', ax=ax2)\n",
    "\n",
    "ax2.set_xlabel('Play Count')\n",
    "ax2.set_ylabel('Artist')\n",
    "ax2.set_title('Bottom-10 Artist ID per play counts')\n",
    "ax2.set_xticks(bottomNArtists['playCount'])\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<div class=\"alert alert-success\" style=\"background-color: AliceBlue; border-color: CornflowerBlue \"> \n",
    "The above plot shows the top 10 artistID's with respect to the play count. This is, to an extent, an indicator of the artist's popularity. Without any assumption about the <b>uniqueness</b> of the artist IDs, the obtained results are very reasonable as the top 10 artists have high play records ranging from <b>1 million to 2.5 million plays</b>.<br>However, merely looking at the top 10 does not give us enough information about the data. Therefore, we also plotted the bottom 10 artists and we noticed again something stange which is that all of the bottom 10 artists have a play count of <b>1</b>. This emphasized what we saw in the second CDF, which is the fact that actually unique artists may be scattered over several artistID's which is causing the play counts to be <b>diluted</b>.<br>\n",
    "<b>Therefore, we need to further inversigate data and verify this artistID's, so that if they are actually not corresponding to unique artists, we should do a cleaning procedure on the data.</b>\n",
    "</div>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Visualizing Artist Play Count Distribution"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.collections.PathCollection at 0x7fdf17e947f0>"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x432 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots()\n",
    "fig.set_size_inches(12, 6)\n",
    "\n",
    "ax.set_xlabel(\"Artist ID\")\n",
    "ax.set_ylabel(\"Play Count\")\n",
    "ax.set_title(\"The Total Number of Play for Every Artist ID\", size = 15)\n",
    "ax.grid(True)\n",
    "ax.scatter(artistPopularity['artistID'],artistPopularity['playCount'], cmap = 'magma', c = artistPopularity['playCount'],  s = 7)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<div class=\"alert alert-success\" style=\"background-color: AliceBlue; border-color: CornflowerBlue \"> \n",
    "The distribution of play counts with the artist ID's shows also that the ID's are somehow scattered and it is also important to point out (which was also observed in the comulative distribution function of artist play counts) is that there is a huge portion of users having very low play counts which are close to 1. Thus, this also emphsizes that there is quite something wrong with the artist ID's which needs to be dealt with later in the notebook.\n",
    "</div>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "All seems clear right now, but ... wait a second! What about the problems indicated above about artist \"disambiguation\"? Are these artist ID we are using referring to unique artists? How can we make sure that such \"opaque\" identifiers point to different bands? Let's try to use some additional dataset to answer this question:  `artist_data.txt` dataset. This time, the schema of the dataset consists in:\n",
    "\n",
    "```\n",
    "artist ID: long int\n",
    "name: string\n",
    "```\n",
    "\n",
    "We will try to find whether a single artist has two different IDs."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### Question 3\n",
    "\n",
    "#### Question 3.1\n",
    "\n",
    "<div class=\"alert alert-info\">\n",
    "Load the data ('\\t'-separated) from `/datasets/lastfm/artist_data.txt` and use the Pandas API to show 5 samples.  \n",
    "</div>\n",
    "\n",
    "**HINT**: `artist_data.txt` has malformed entries. If you encounter some error when parsing lines in data because of invalid entries, functions `pandas.DataFrame.dropna()` and `pandas.to_numeric()` will help you to eliminate these entries.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.5/dist-packages/IPython/core/interactiveshell.py:2785: DtypeWarning: Columns (0,1) have mixed types. Specify dtype option on import or set low_memory=False.\n",
      "  interactivity=interactivity, compiler=compiler, result=result)\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>artistID</th>\n",
       "      <th>name</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1134999</td>\n",
       "      <td>06Crazy Life</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>6821360</td>\n",
       "      <td>Pang Nakarin</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>10113088</td>\n",
       "      <td>Terfel, Bartoli- Mozart: Don</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>10151459</td>\n",
       "      <td>The Flaming Sidebur</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>6826647</td>\n",
       "      <td>Bodenstandig 3000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   artistID                          name\n",
       "0   1134999                  06Crazy Life\n",
       "1   6821360                  Pang Nakarin\n",
       "2  10113088  Terfel, Bartoli- Mozart: Don\n",
       "3  10151459           The Flaming Sidebur\n",
       "4   6826647             Bodenstandig 3000"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Read from the .csv file\n",
    "artistDF = pd.read_csv(base + \"artist_data.txt\",\n",
    "                       sep = \"\\t\",\n",
    "                       header = None,\n",
    "                       names=['artistID', 'name'])\n",
    "\n",
    "# Filter out possible malformed entries\n",
    "\n",
    "# Change the ID's that can't be converted to NA\n",
    "artistDF['artistID'] = pd.to_numeric(artistDF['artistID'], errors = 'coerce')\n",
    "\n",
    "# Drop the rows containing NA\n",
    "artistDF = artistDF.dropna()\n",
    "\n",
    "# Cast each column to its datatype\n",
    "artistDF['name'] = artistDF['name'].astype(str, errors='ignore')\n",
    "artistDF['artistID'] = artistDF['artistID'].astype(np.int64, errors='ignore')\n",
    "\n",
    "\n",
    "artistDF[:5]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<div class=\"alert alert-success\" style=\"background-color: AliceBlue; border-color: CornflowerBlue \"> \n",
    "We can start to see directly that some of the names of the artists have some typos for <b>\"The Flaming Sideburns\"</b> and some contain song names with the track number like <b>\"06Crazy Life\"</b>. Hence, our speculation is looking more and more consistent and relevant as artistID don't only represent merely artist names. To make it more clear, let's try to search for some famous artist in the artist name.\n",
    "</div>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "#### Question 3.2\n",
    "<div class=\"alert alert-info\">\n",
    "Find 20 artists whose name contains `Aerosmith`. Take a look at artists that have ID equal to `1000010` and `2082323`. In your opinion, are they pointing to the same artist?  \n",
    "</div>\n",
    "\n",
    "**HINT**: Function `contains(string)` can be useful in this case."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>artistID</th>\n",
       "      <th>name</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>9925</th>\n",
       "      <td>10586006</td>\n",
       "      <td>Dusty Springfield/Aerosmith</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12010</th>\n",
       "      <td>6946007</td>\n",
       "      <td>Aerosmith/RunDMC</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12456</th>\n",
       "      <td>10475683</td>\n",
       "      <td>Aerosmith: Just Push Play</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14567</th>\n",
       "      <td>1083031</td>\n",
       "      <td>Aerosmith/ G n R</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15761</th>\n",
       "      <td>6872848</td>\n",
       "      <td>Britney, Nsync, Nelly, Aerosmith,Mary J Blige.</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19409</th>\n",
       "      <td>10586963</td>\n",
       "      <td>Green Day - Oasis - Eminem - Aerosmith</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20337</th>\n",
       "      <td>10028830</td>\n",
       "      <td>The Aerosmith Antology2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25166</th>\n",
       "      <td>10300357</td>\n",
       "      <td>Run-DMC + Aerosmith</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26399</th>\n",
       "      <td>2027746</td>\n",
       "      <td>Aerosmith by MusicInter.com</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>28469</th>\n",
       "      <td>1140418</td>\n",
       "      <td>[rap]Run DMC and Aerosmith</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30207</th>\n",
       "      <td>10237208</td>\n",
       "      <td>Aerosmith + Run DMC</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>33127</th>\n",
       "      <td>10588537</td>\n",
       "      <td>Aerosmith, Kid Rock, &amp; Run DMC</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>41891</th>\n",
       "      <td>9934757</td>\n",
       "      <td>Aerosmith - Big Ones</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>42294</th>\n",
       "      <td>10437510</td>\n",
       "      <td>Green Day ft. Oasis &amp; Aerosmith</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>48005</th>\n",
       "      <td>6936680</td>\n",
       "      <td>RUN DNC &amp; Aerosmith</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>49323</th>\n",
       "      <td>10479781</td>\n",
       "      <td>Aerosmith Hits</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>55799</th>\n",
       "      <td>10114147</td>\n",
       "      <td>Charlies Angels - Aerosmith</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>62947</th>\n",
       "      <td>1262439</td>\n",
       "      <td>Kid Rock, Run DMC &amp; Aerosmith</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>63907</th>\n",
       "      <td>7032554</td>\n",
       "      <td>Aerosmith &amp; Run-D.M.C.</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>64842</th>\n",
       "      <td>10033592</td>\n",
       "      <td>Aerosmith?</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       artistID                                            name\n",
       "9925   10586006                     Dusty Springfield/Aerosmith\n",
       "12010   6946007                                Aerosmith/RunDMC\n",
       "12456  10475683                       Aerosmith: Just Push Play\n",
       "14567   1083031                                Aerosmith/ G n R\n",
       "15761   6872848  Britney, Nsync, Nelly, Aerosmith,Mary J Blige.\n",
       "19409  10586963          Green Day - Oasis - Eminem - Aerosmith\n",
       "20337  10028830                         The Aerosmith Antology2\n",
       "25166  10300357                             Run-DMC + Aerosmith\n",
       "26399   2027746                     Aerosmith by MusicInter.com\n",
       "28469   1140418                      [rap]Run DMC and Aerosmith\n",
       "30207  10237208                             Aerosmith + Run DMC\n",
       "33127  10588537                  Aerosmith, Kid Rock, & Run DMC\n",
       "41891   9934757                            Aerosmith - Big Ones\n",
       "42294  10437510                 Green Day ft. Oasis & Aerosmith\n",
       "48005   6936680                             RUN DNC & Aerosmith\n",
       "49323  10479781                                  Aerosmith Hits\n",
       "55799  10114147                     Charlies Angels - Aerosmith\n",
       "62947   1262439                   Kid Rock, Run DMC & Aerosmith\n",
       "63907   7032554                          Aerosmith & Run-D.M.C.\n",
       "64842  10033592                                      Aerosmith?"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# get artists whose name contains \"Aerosmith\"\n",
    "artistDF[artistDF.name.str.contains('Aerosmith')][:20]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<div class=\"alert alert-success\" style=\"background-color: AliceBlue; border-color: CornflowerBlue \"> \n",
    "It is evident now that actually several artistID's can correspond to one unique artist. This is causing inconsistency in our dataset which needs to be dealt with to achieve better more consistent semantics.\n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>artistID</th>\n",
       "      <th>name</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>958509</th>\n",
       "      <td>1000010</td>\n",
       "      <td>Aerosmith</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        artistID       name\n",
       "958509   1000010  Aerosmith"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "artistDF[artistDF['artistID'] == 1000010]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>artistID</th>\n",
       "      <th>name</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1457634</th>\n",
       "      <td>2082323</td>\n",
       "      <td>01 Aerosmith</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "         artistID          name\n",
       "1457634   2082323  01 Aerosmith"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "artistDF[artistDF['artistID'] == 2082323]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<div class=\"alert alert-success\" style=\"background-color: AliceBlue; border-color: CornflowerBlue \">\n",
    "The two artistID's 1000010 and 2082323 both correspond to Aerosmith.\n",
    "</div>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "To answer this question correctly, we need to use an additional dataset `artist_alias.txt` which contains the ids of mispelled artists and standard artists. The schema of the dataset consists in:\n",
    "\n",
    "```\n",
    "mispelledID ID: long int\n",
    "standard ID: long int\n",
    "```"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "#### Question 3.3\n",
    "<div class=\"alert alert-info\">\n",
    "Using Pandas, load the dataset from `/datasets/lastfm/artist_alias.txt` then show 5 samples.\n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "scrolled": false,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>misspelledArtistID</th>\n",
       "      <th>standardArtistID</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1092764</td>\n",
       "      <td>1000311</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1095122</td>\n",
       "      <td>1000557</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>6708070</td>\n",
       "      <td>1007267</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>10088054</td>\n",
       "      <td>1042317</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1195917</td>\n",
       "      <td>1042317</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   misspelledArtistID  standardArtistID\n",
       "0             1092764           1000311\n",
       "1             1095122           1000557\n",
       "2             6708070           1007267\n",
       "3            10088054           1042317\n",
       "4             1195917           1042317"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "artistAliasDF = pd.read_csv(base + 'artist_alias.txt',\n",
    "                            sep = '\\t',\n",
    "                            header = None,\n",
    "                            names=['misspelledArtistID', 'standardArtistID'])\n",
    "\n",
    "# Filter out possible malformed entries\n",
    "\n",
    "# Change the ID's that can't be converted NA\n",
    "artistAliasDF['misspelledArtistID'] = pd.to_numeric(artistAliasDF['misspelledArtistID'], errors='coerce')\n",
    "artistAliasDF['standardArtistID'] = pd.to_numeric(artistAliasDF['standardArtistID'], errors='coerce')\n",
    "\n",
    "# Drop the rows containing NA\n",
    "artistAliasDF = artistAliasDF.dropna()\n",
    "\n",
    "# Cast each column to its datatype\n",
    "artistAliasDF['misspelledArtistID'] = artistAliasDF['misspelledArtistID'].astype(np.int64, errors='ignore')\n",
    "artistAliasDF['standardArtistID'] = artistAliasDF['standardArtistID'].astype(np.int64, errors='ignore')\n",
    "\n",
    "# Show 5 samples\n",
    "artistAliasDF[:5]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<div class=\"alert alert-success\" style=\"background-color: AliceBlue; border-color: CornflowerBlue \"> \n",
    "We can see that the artistID's 10088054 and 1195917 actually correspond to the standard artist ID 1042317.\n",
    "</div>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "#### Question 3.4\n",
    "<div class=\"alert alert-info\">\n",
    "Verify the answer of question 3.2 (\"Are artists that have ID equal to `1000010` and `2082323` the same ?\") by finding the standard ids corresponding to the mispelled ids `1000010` and `2082323` respectively.\n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>misspelledArtistID</th>\n",
       "      <th>standardArtistID</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1135</th>\n",
       "      <td>2082323</td>\n",
       "      <td>1000010</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      misspelledArtistID  standardArtistID\n",
       "1135             2082323           1000010"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "artistAliasDF[artistAliasDF.misspelledArtistID == 2082323]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>misspelledArtistID</th>\n",
       "      <th>standardArtistID</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "Empty DataFrame\n",
       "Columns: [misspelledArtistID, standardArtistID]\n",
       "Index: []"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "artistAliasDF[artistAliasDF.misspelledArtistID == 1000010]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<div class=\"alert alert-success\" style=\"background-color: AliceBlue; border-color: CornflowerBlue \"> \n",
    "The results obtained in <b>3.2</b> can be verified here where the consistent artistID for <b>AeroSmith</b> is 1000010 and 2082323 actually corresponds to Aerosmith (artistID = 1000010).<br>\n",
    "<b>NOTE:</b> the consistent artistID's do not exist among the misspelledArtistID column entries in the dataframe.\n",
    "</div>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### Question 4\n",
    "\n",
    "The mispelled or nonstandard information about artist make our results in the previous queries a bit \"sloppy\". To overcome this problem, we can replace all mispelled artist ids by the corresponding standard ids and re-compute the basic descriptive statistics on the \"amended\" data.\n",
    "\n",
    "#### Question 4.2\n",
    "<div class=\"alert alert-info\">\n",
    "Replace the non-standard artist ids in the dataframe that was loaded from `/datasets/lastfm/user_artist_data.txt` by the corresponding standard ids then show 5 samples.\n",
    "</div>\n",
    "\n",
    "\n",
    "**NOTE 1**: If an id doesn't exist in the dictionary as a mispelled id, it is really a standard id.\n",
    "\n",
    "**NOTE 2**: be careful! you need to be able to verify that you indeed solved the problem of having bad artist IDs. In principle, for the new data to be correct, we should to have duplicate pairs (user, artist), potentially with different play counts, right? In answering the question, please **show** that you indeed fixed the problem and find a way to fix the duplicates (user, artist).\n",
    "\n",
    "**HINT**: have a look to the functions `pandas.merge()` and `pandas.Dataframe.combine_first()`."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Cleaning the dataset took 26.18 seconds\n"
     ]
    }
   ],
   "source": [
    "\n",
    "t0 = time()\n",
    "\n",
    "# Join the two dataframes based on the misspelledArtistID\n",
    "newUserArtistDF = pd.merge(userArtistDF, artistAliasDF, left_on='artistID', right_on='misspelledArtistID', how='left')\n",
    "\n",
    "# Get the final realArtistID\n",
    "'''\n",
    "    Take into consideration that the standardArtistID column \n",
    "    will not take a value after the join in case where the artistID \n",
    "    was originally correct\n",
    "'''\n",
    "newUserArtistDF['realArtistID'] = newUserArtistDF['standardArtistID'].combine_first(newUserArtistDF['artistID']).astype(np.int64)\n",
    "\n",
    "# Delete columns not required\n",
    "newUserArtistDF.drop(['artistID', 'standardArtistID', 'misspelledArtistID'], axis=1, inplace=True)\n",
    "\n",
    "# Fix duplicates of the pair (userID, realArtistID) by summing the play counts of identical artist-user-pairs\n",
    "newUserArtistDF = newUserArtistDF.groupby(['userID','realArtistID'], as_index=False).sum()\n",
    "\n",
    "print('Cleaning the dataset took %.2f seconds' % (time() - t0))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "Int64Index: 24133304 entries, 0 to 24133303\n",
      "Data columns (total 3 columns):\n",
      "userID          int64\n",
      "realArtistID    int64\n",
      "playCount       int32\n",
      "dtypes: int32(1), int64(2)\n",
      "memory usage: 644.4 MB\n"
     ]
    }
   ],
   "source": [
    "# Print some details on the cleaned data\n",
    "newUserArtistDF.info()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>userID</th>\n",
       "      <th>realArtistID</th>\n",
       "      <th>playCount</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>90</td>\n",
       "      <td>24</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>90</td>\n",
       "      <td>46</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>90</td>\n",
       "      <td>121</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>90</td>\n",
       "      <td>151</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>90</td>\n",
       "      <td>227</td>\n",
       "      <td>16</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   userID  realArtistID  playCount\n",
       "0      90            24          1\n",
       "1      90            46          4\n",
       "2      90           121          1\n",
       "3      90           151          1\n",
       "4      90           227         16"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# print the first 5 records\n",
    "newUserArtistDF[:5]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "# test the dataframe obtained to check whether all the records and duplicates are fixed\n",
    "assert (newUserArtistDF.groupby(['userID', 'realArtistID'], as_index=False).count()['playCount'] == 1).all(),  \\\n",
    "    'Each combination user/artist should appear only once'"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<div class=\"alert alert-success\" style=\"background-color: AliceBlue; border-color: CornflowerBlue \"> \n",
    "There is no printed message, so the assetion is correct and we have only unique artist-user-pairs. Therefore, our cleaning process is successful.\n",
    "</div>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### Question 5\n",
    "\n",
    "Well, our data frame contains clean and \"standard\" data. We can use it to redo previous statistic queries. Now that we have all data, including the relation (`artistID`/`name`), please find a way to use the artist name instead of its ID in all required plots.\n",
    "\n",
    "#### Question 5.1\n",
    "<div class=\"alert alert-info\">\n",
    "How many unique artists? Compare with the result when using old data.\n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Total n. of artists:  1568126\n"
     ]
    }
   ],
   "source": [
    "uniqueArtists = newUserArtistDF['realArtistID'].nunique()\n",
    "\n",
    "print(\"Total n. of artists: \", uniqueArtists)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<div class=\"alert alert-success\" style=\"background-color: AliceBlue; border-color: CornflowerBlue \"> \n",
    "The number of unique artists obtained now (1568126 artists) after cleaning is less than the number of artistID's obtained at first (1631028 artists) by about <b> 63K artist </b> which is a big deal. So the data cleaning process was essential in our work for achieving better results.\n",
    "    </div>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "#### Question 5.2\n",
    "\n",
    "<div class=\"alert alert-info\">\n",
    "Who are the top-10 artistis?\n",
    "<ul>\n",
    "  <li>In terms of absolute play counts</li>\n",
    "  <li>In terms of \"audience size\", that is, how many users listened to one of their track at least once</li>\n",
    "</ul>  \n",
    "\n",
    "Plot the results, and explain the figures you obtain.\n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5,1,'Top-10 Artist per play counts')"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x432 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# calculate top-10 artists in term of play counts\n",
    "topN = 10\n",
    "\n",
    "# Group the artists by their ID's and aggregate by summing the play counts\n",
    "artistPopularity = newUserArtistDF.groupby(['realArtistID'], as_index = False)['playCount'].sum()\n",
    "\n",
    "# Get the name of the artists based on their IDs\n",
    "artistPopularity = pd.merge(artistPopularity, artistDF, right_on='artistID', left_on='realArtistID', how='left')\n",
    "\n",
    "# Get the top 10 and the bottom 10 artists with respect to the total play count\n",
    "rankedArtist = artistPopularity.sort_values('playCount', ascending=False)\n",
    "\n",
    "# Get the top 10 artists\n",
    "topNArtists = rankedArtist[:topN]\n",
    "\n",
    "fig, ax = plt.subplots()\n",
    "\n",
    "fig.set_size_inches(10,6)\n",
    "\n",
    "topNArtists['playCount'].plot(kind='barh', ax=ax)\n",
    "\n",
    "ax.set_yticklabels(zip(topNArtists['name'],topNArtists['artistID']))\n",
    "ax.set_xlabel('Play Count')\n",
    "ax.set_title('Top-10 Artist per play counts')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<div class=\"alert alert-success\" style=\"background-color: AliceBlue; border-color: CornflowerBlue \">\n",
    "The top 10 artists with respect to the play count are identical those obtained before data cleaning. This means mostly the inconsistent ID's do not contribute largely in play counts of the artists which have high records but rather with those with the artists with smaller records.\n",
    "</div>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<div class=\"alert alert-success\" style=\"background-color: AliceBlue; border-color: CornflowerBlue \">\n",
    "Let's now look at the bottom 10 artists.\n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>name</th>\n",
       "      <th>playCount</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>724712</th>\n",
       "      <td>ファニー</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1186250</th>\n",
       "      <td>The Guess Who w/ Chad Allen</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1186249</th>\n",
       "      <td>倖月 美和</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>89820</th>\n",
       "      <td>Ultra Lounge Series</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>724714</th>\n",
       "      <td>アバンギャルド</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1186245</th>\n",
       "      <td>VA - Round Four</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>724715</th>\n",
       "      <td>Cherrystone</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1186243</th>\n",
       "      <td>Sykeste Folkan</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1186242</th>\n",
       "      <td>Sunn O))) - Candlewolf Of The Golden Chalice</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1186241</th>\n",
       "      <td>Jaime 'Rantam' López</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1568125</th>\n",
       "      <td>NaN</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                 name  playCount\n",
       "724712                                           ファニー          1\n",
       "1186250                   The Guess Who w/ Chad Allen          1\n",
       "1186249                                         倖月 美和          1\n",
       "89820                             Ultra Lounge Series          1\n",
       "724714                                        アバンギャルド          1\n",
       "1186245                               VA - Round Four          1\n",
       "724715                                    Cherrystone          1\n",
       "1186243                                Sykeste Folkan          1\n",
       "1186242  Sunn O))) - Candlewolf Of The Golden Chalice          1\n",
       "1186241                          Jaime 'Rantam' López          1\n",
       "1568125                                           NaN          1"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Let's also display the bottom 10 artists with respect to play count\n",
    "rankedArtist[['name','playCount']][-topN-1:]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<div class=\"alert alert-success\" style=\"background-color: AliceBlue; border-color: CornflowerBlue \">\n",
    "We still have many artists with 1 as a play count but those artists are unique even though some have names that do not exist in the dataset.</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "slideshow": {
     "slide_type": "-"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x432 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# calculate top-10 artists in term of audience size\n",
    "\n",
    "topN = 10\n",
    "\n",
    "artistAudience = newUserArtistDF.groupby(['realArtistID'], as_index = False)['userID'].count()\n",
    "artistAudience = pd.merge(artistAudience, artistDF, right_on='artistID', left_on='realArtistID', how='left')\n",
    "topNArtist = artistAudience.sort_values('userID', ascending=False)[:topN]\n",
    "\n",
    "fig, ax = plt.subplots()\n",
    "\n",
    "fig.set_size_inches(12,6)\n",
    "\n",
    "topNArtist['userID'].plot(kind='barh', ax=ax)\n",
    "\n",
    "ax.set_yticklabels(zip(topNArtist['name'],topNArtist['artistID']))\n",
    "ax.set_xlabel('Audience Size')\n",
    "ax.set_title('Top-10 Artist in tems of Audience Size')\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<div class=\"alert alert-success\" style=\"background-color: AliceBlue; border-color: CornflowerBlue \">\n",
    "After displaying the top 10 artists in terms of the audience size, we can see than they are actually different from the top 10 artists with respect to absolute play counts. Hence, the <b>popularity invesitgation</b> is dependent on how we deal with it; whether its through play counts or through audience size. For example, <b>Coldplay</b> did not appear in the top 10 artist play counts but has appeared here. Thus, it has been important to differentiate between the <b>total play count</b> and the <b>audience size</b> in the same way that notebook previously discusses the difference between the <b>curiousity of the user</b> and his/her <b>absolute play count</b> because we cannot conclude one from the others. The reason is due to the fact that many users actually play tracks for a single artists a massive number of times which leads to a difference between the two aproaches.\n",
    "</div>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "#### Question 5.3\n",
    "\n",
    "<div class=\"alert alert-info\">\n",
    "Who are the top-10 users?\n",
    "<ul>\n",
    "  <li>In terms of absolute play counts</li>\n",
    "  <li>In terms of \"curiosity\", that is, how many different artists they listened to</li>\n",
    "\n",
    "</ul>  \n",
    "\n",
    "Plot the results\n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x432 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# calculate top-10 user in term of play counts\n",
    "topN = 10\n",
    "\n",
    "userPlayCount = newUserArtistDF.groupby('userID', as_index=False)['playCount'].sum()\n",
    "rankedUsers = userPlayCount.sort_values('playCount', ascending = False)\n",
    "topNUser = rankedUsers[:topN]\n",
    "\n",
    "fig, ax = plt.subplots()\n",
    "\n",
    "fig.set_size_inches(12,6)\n",
    "\n",
    "topNUser['playCount'].plot(kind='barh', ax=ax)\n",
    "\n",
    "ax.set_yticklabels(topNUser['userID'])\n",
    "ax.set_xlabel('Play Count')\n",
    "ax.set_ylabel('userID')\n",
    "ax.set_title('Top-10 User per play count')\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x432 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# calculate top-10 user in term of curiosity\n",
    "topN = 10\n",
    "\n",
    "userCuriosity = newUserArtistDF.groupby('userID', as_index=False)['realArtistID'].count()\n",
    "rankedUsersCuriosity = userCuriosity.sort_values('realArtistID', ascending = False)\n",
    "topNUserCuriosity = rankedUsersCuriosity[:topN]\n",
    "fig, ax = plt.subplots()\n",
    "\n",
    "fig.set_size_inches(12,6)\n",
    "\n",
    "topNUserCuriosity['realArtistID'].plot(kind='barh', ax=ax)\n",
    "\n",
    "ax.set_yticklabels(topNUserCuriosity['userID'])\n",
    "ax.set_xlabel('Play Count')\n",
    "ax.set_title('Top-10 User per curiosity')\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<div class=\"alert alert-success\" style=\"background-color: AliceBlue; border-color: CornflowerBlue \">\n",
    "Above, we obtained the results for the top 10 users in terms of <b>total play count</b> and <b>user curiousity</b>. We can clearly see that their is <b>no apparent correlation</b> between the two results. The most curious userID 2030067 doesn't even appear in the top 10 users in terms of play count. Therefore, it is also important to differetiate between the two measures and this also verifies our different results between the <b>top 10 play count artists</b> and the <b>top 10 audience size artists</b>. And so it is important also to take into consideration that indeed many users actually play the same music a high number of times in such a way that they disappear from the top 10 curious users.\n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 104,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The number of users with a play count higher than 50K 151\n",
      "The 0.999th quantile of the play count per user is 39550.69700000045\n"
     ]
    }
   ],
   "source": [
    "print(\"The number of users with a play count higher than 50K\", userPlayCount[userPlayCount.playCount > 50000].count().values[0])\n",
    "users2BRemoved = userPlayCount[userPlayCount.playCount > 50000]['userID'].values\n",
    "userPlayCount = userPlayCount[userPlayCount.playCount <= 50000].sort_values('playCount', ascending = False)\n",
    "userCuriosity = userCuriosity.drop(userCuriosity.loc[userCuriosity.userID.isin(users2BRemoved)].index)\n",
    "print(\"The 0.999th quantile of the play count per user is\", newUserArtistDF.groupby('userID', as_index=False)[\"playCount\"].sum().quantile(0.999).values[1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 91,
   "metadata": {},
   "outputs": [],
   "source": [
    "newUserArtistDF = newUserArtistDF.drop(newUserArtistDF.loc[newUserArtistDF.userID.isin(users2BRemoved)].index)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now we have some valuable information about the data. It's the time to study how to build a statistical models."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "# 2. Build a statistical models to make recommendations\n",
    "\n",
    "## 2.1 Introduction to recommender systems\n",
    "\n",
    "In a recommendation-system application there are two classes of entities, which we shall refer to as `users` and `items`. Users have preferences for certain items, and these preferences must be inferred from the data. The data itself is represented as a `preference matrix` $A$, giving for each user-item pair, a value that represents what is known about the degree of preference of that user for that item. The table below is an example for a `preference matrix` of 5 users and `k` items. The `preference matrix` is also known as `utility matrix`.\n",
    "\n",
    "| .  | IT1 | IT2 | IT3 | ... | ITk |\n",
    "|----|---|---|---|---|---|\n",
    "| U1 | 1 | na | 5 | ... | 3 |\n",
    "| U2 | na | 2 | na | ... | 2 |\n",
    "| U3 | 5 | na | 3 | ... | na |\n",
    "| U4 | 3 | 3 | na | ... | 4 |\n",
    "| U5 | na | 1 | na | ... | ... |\n",
    "\n",
    "The value of row i, column j expresses how much does user `i` like item `j`. The values are often the rating scores of users for items. An unknown value implies that we have no explicit information about the user's preference for the item. The goal of a recommendation system is to predict \"the blanks\" in the `preference matrix`. For example, assume that the rating score is from 1 (dislike) to 5 (love), would user `U5` like `IT3` ? We have two approaches:\n",
    "\n",
    "* Designing our recommendation system to take into account properties of items such as brand, category, price... or even the similarity of their names. We can denote the similarity of items `IT2` and `IT3`, and then conclude that because user `U5` did not like `IT2`, they were unlikely to enjoy SW2 either.\n",
    "\n",
    "* We might observe that the people who rated both `IT2` and `IT3` tended to give them similar ratings. Thus, we could conclude that user `U5` would also give `IT3` a low rating, similar to `U5`'s rating of `IT2`\n",
    "\n",
    "It is not necessary to predict every blank entry in a `utility matrix`. Rather, it is only necessary to discover some entries in each row that are likely to be high. In most applications, the recommendation system does not oﬀer users a ranking of all items, but rather suggests a few that the user should value highly. It may not even be necessary to ﬁnd all items with the highest expected ratings, but only to ﬁnd a large subset of those with the highest ratings.\n",
    "\n",
    "\n",
    "## 2.2 Families of recommender systems\n",
    "\n",
    "In general, recommender systems can be categorized into two groups:\n",
    "\n",
    "* **Content-Based** systems focus on properties of items. Similarity of items is determined by measuring the similarity in their properties.\n",
    "\n",
    "* **Collaborative-Filtering** systems focus on the relationship between users and items. Similarity of items is determined by the similarity of the ratings of those items by the users who have rated both items.\n",
    "\n",
    "In the usecase of this notebook, artists take the role of `items`, and `users` keep the same role as `users`.\n",
    "Since we have no information about `artists`, except their names, we cannot build a `content-based` recommender system.\n",
    "\n",
    "Therefore, in the rest of this notebook, we only focus on `Collaborative-Filtering` algorithms.\n",
    "\n",
    "## 2.3 Collaborative-Filtering \n",
    "In this section, we study a member of a broad class of algorithms called `latent-factor` models. They try to explain observed interactions between large numbers of users and products through a relatively small number of unobserved, underlying reasons. It is analogous to explaining why millions of people buy a particular few of thousands of possible albums by describing users and albums in terms of tastes for perhaps tens of genres, tastes which are **not directly observable or given** as data. \n",
    "\n",
    "First, we formulate the learning problem as a matrix completion problem. Then, we will use a type of `matrix factorization` model to \"fill in\" the blanks.  We are given implicit ratings that users have given certain items (that is, the number of times they played a particular artist) and our goal is to predict their ratings for the rest of the items. Formally, if there are $n$ users and $m$ items, we are given an $n \\times m$ matrix $R$ in which the generic entry $(u, i)$ represents the rating for item $i$ by user $u$. **Matrix $R$ has many missing entries indicating unobserved ratings, and our task is to estimate these unobserved ratings**.\n",
    "\n",
    "A popular approach to the matrix completion problem is **matrix factorization**, where we want to \"summarize\" users and items with their **latent factors**.\n",
    "\n",
    "### 2.3.1 Basic idea and an example of Matrix Factorization\n",
    "For example, given a preference matrix 5x5 as below, we want to approximate this matrix into the product of two smaller matrixes $X$ and $Y$ .\n",
    "\n",
    "$$\n",
    "M = \n",
    "\\begin{bmatrix}\n",
    " 5 & 2 & 4 & 4 & 3 \\\\\n",
    " 3 & 1 & 2 & 4 & 1 \\\\\n",
    " 2 &  & 3 & 1 & 4 \\\\\n",
    " 2 & 5 & 4 & 3 & 5 \\\\\n",
    " 4 & 4 & 5 & 4 &  \\\\\n",
    "\\end{bmatrix}\n",
    "\\approx M^\\prime =\n",
    "\\begin{bmatrix}\n",
    " x_{11} & x_{12} \\\\\n",
    " x_{21} & x_{22} \\\\\n",
    " x_{31} & x_{32} \\\\\n",
    " x_{41} & x_{42} \\\\\n",
    " x_{51} & x_{52} \\\\\n",
    "\\end{bmatrix}\n",
    "\\times\n",
    "\\begin{bmatrix}\n",
    " y_{11} & y_{12} & y_{13} & y_{14} & y_{15} \\\\\n",
    " y_{21} & y_{22} & y_{23} & y_{24} & y_{25} \\\\\n",
    "\\end{bmatrix}\n",
    "$$\n",
    "\n",
    "$M^\\prime$ is an approximation that is as close to A as possible. To calculate how far from $M$ $M^\\prime$ is, we often calculate the sum of squared distances of non-empty elements in $M$ and the corresponding elements in $M^\\prime$.\n",
    "In this way, for $M^\\prime$, besides the approximated elements in $M$, we also have the non-observed elements. Therefore, to see how much does user `i` like item `j`, we simply pick up the value of $M^\\prime_{i,j}$.\n",
    "\n",
    "The challenge is how to calculate $X$ and $Y$. The bad news is that this can't be solved directly for both the best $X$ and best $Y$ at the same time. Fortunately, if $Y$ is known, we can calculate the best of $X$, and vice versa. It means from the initial values of $X$ and $Y$ in the beginning, we calculate the best $X$ according to $Y$, and then calculate the best $Y$ according to the new $X$. This process is repeated until the distance from $XY$ to $M$ is small. It's simple, right ?\n",
    "\n",
    "Let's take an example. To compute the approximation for the above 5x5 matrix $M$, first, we initialize the value of $X$ and $Y$ as below.\n",
    "\n",
    "$$\n",
    "M^\\prime = X \\times Y =\n",
    "\\begin{bmatrix}\n",
    " 1 & 1 \\\\\n",
    " 1 & 1 \\\\\n",
    " 1 & 1 \\\\\n",
    " 1 & 1 \\\\\n",
    " 1 & 1 \\\\\n",
    "\\end{bmatrix}\n",
    "\\times\n",
    "\\begin{bmatrix}\n",
    " 1 & 1 & 1 & 1 & 1 \\\\\n",
    " 1 & 1 & 1 & 1 & 1 \\\\\n",
    "\\end{bmatrix}\n",
    "=\n",
    "\\begin{bmatrix}\n",
    " 2 & 2 & 2 & 2 & 2 \\\\\n",
    " 2 & 2 & 2 & 2 & 2 \\\\\n",
    " 2 & 2 & 2 & 2 & 2 \\\\\n",
    " 2 & 2 & 2 & 2 & 2 \\\\\n",
    " 2 & 2 & 2 & 2 & 2 \\\\\n",
    "\\end{bmatrix}\n",
    "$$\n",
    "\n",
    "With the initial iteration, we calculate the the Root-Mean-Square Error from $XY$ to $M$.\n",
    "\n",
    "Consider the ﬁrst rows of $M$ and $XY$ . We subtract the first row of $XY$ from the entries in the ﬁrst row of $M$, to get $3,0,2,2,1$. We square and sum these to get $18$. \n",
    "\n",
    "In the second row, we do the same to get $1,−1,0,2,−1$, square and sum to get $7$. \n",
    "\n",
    "In the third row, the second column is blank, so that entry is ignored when computing the RMSE. The diﬀerences are $0,1,−1,2$ and the sum of squares is $6$. \n",
    "\n",
    "For the fourth row, the diﬀerences are $0,3,2,1,3$ and the sum of squares is $23$. \n",
    "\n",
    "The ﬁfth row has a blank entry in the last column, so the diﬀerences are $2,2,3,2$ and the sum of squares is $21$. \n",
    "\n",
    "When we sum the sums from each of the ﬁve rows, we get $18+7+6+23+21 = 75$. So, $RMSE=\\sqrt{75/23}=1.806$ where $23$ is the number of non-empty values in $M$.\n",
    "\n",
    "Next, with the given value of $Y$, we calculate $X$ by finding the best value for $X_{11}$.\n",
    "\n",
    "$$\n",
    "M^\\prime = X \\times Y =\n",
    "\\begin{bmatrix}\n",
    " x & 1 \\\\\n",
    " 1 & 1 \\\\\n",
    " 1 & 1 \\\\\n",
    " 1 & 1 \\\\\n",
    " 1 & 1 \\\\\n",
    "\\end{bmatrix}\n",
    "\\times\n",
    "\\begin{bmatrix}\n",
    " 1 & 1 & 1 & 1 & 1 \\\\\n",
    " 1 & 1 & 1 & 1 & 1 \\\\\n",
    "\\end{bmatrix}\n",
    "=\n",
    "\\begin{bmatrix}\n",
    " x+1 & x+1 & x+1 & x+1 & x+1 \\\\\n",
    " 2 & 2 & 2 & 2 & 2 \\\\\n",
    " 2 & 2 & 2 & 2 & 2 \\\\\n",
    " 2 & 2 & 2 & 2 & 2 \\\\\n",
    " 2 & 2 & 2 & 2 & 2 \\\\\n",
    "\\end{bmatrix}\n",
    "$$\n",
    "\n",
    "Now, to minimize the $RMSE$  we minimize the difference of the first rows $(5−(x+1))^2 + (2−(x+1))^2 + (4−(x+1))^2 + (4−(x+1))^2 + (3−(x+1))^2$. By  taking the derivative and set that equal to 0, we pick $x=2.6$\n",
    "\n",
    "Given the new value of $X$, we can calculate the best value for $Y$.\n",
    "\n",
    "$$\n",
    "M^\\prime = X \\times Y =\n",
    "\\begin{bmatrix}\n",
    " 2.6 & 1 \\\\\n",
    " 1 & 1 \\\\\n",
    " 1 & 1 \\\\\n",
    " 1 & 1 \\\\\n",
    " 1 & 1 \\\\\n",
    "\\end{bmatrix}\n",
    "\\times\n",
    "\\begin{bmatrix}\n",
    " y & 1 & 1 & 1 & 1 \\\\\n",
    " 1 & 1 & 1 & 1 & 1 \\\\\n",
    "\\end{bmatrix}\n",
    "=\n",
    "\\begin{bmatrix}\n",
    " 3.6 & 3.6 & 3.6 & 3.6 & 3.6 \\\\\n",
    " 2 & 2 & 2 & 2 & 2 \\\\\n",
    " 2 & 2 & 2 & 2 & 2 \\\\\n",
    " 2 & 2 & 2 & 2 & 2 \\\\\n",
    " 2 & 2 & 2 & 2 & 2 \\\\\n",
    "\\end{bmatrix}\n",
    "$$\n",
    "\n",
    "By doing the same process as before, we can pick value for $y=1.617$. After that, we can check if the $RMSE$ is not converged, we continue to update $X$ by $Y$ and vice versa. In this example, for simple, we only update one element of each matrix in each iteration. In practice, we can update a full row or full matrix at once.\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 2.3.2 Matrix Factorization: Objective and ALS Algorithm\n",
    "\n",
    "More formally, in general, we select $k$ latent features, and describe each user $u$ with a $k-$dimensional vector $x_u$, and each item $i$ with a $k-$dimensional vector $y_i$.\n",
    "\n",
    "Then, to predict user $u$'s rating for item $i$, we do as follows: $ r_{ui} \\approx x_{u}^{T}y_i$.\n",
    "\n",
    "This can be put, more elegantly, in a matrix form. Let $x_1, \\cdots x_n \\in \\mathbb{R}^k$ be the factors for the users, and $y_1, \\cdots y_m \\in \\mathbb{R}^k$ the factors for the items. The $k \\times n$ user matrix $X$ and the $k \\times m$ item matrix $Y$ are then defined by:\n",
    "\n",
    "$$\n",
    "X = \n",
    "\\begin{bmatrix}\n",
    " |   &         & |  \\\\\n",
    "x_1  &  \\cdots & x_n\\\\\n",
    " |   &         & |  \\\\\n",
    "\\end{bmatrix}\n",
    "$$\n",
    "\n",
    "$$\n",
    "Y = \n",
    "\\begin{bmatrix}\n",
    " |   &         & |  \\\\\n",
    "y_1  &  \\cdots & y_i\\\\\n",
    " |   &         & |  \\\\\n",
    "\\end{bmatrix}\n",
    "$$\n",
    "\n",
    "Our goal is to estimate the complete ratings matrix $R \\approx X^{T} Y$. We can formulate this problem as an optimization problem in which we aim to minimize an objective function and find optimal $X$ and $Y$ . In particular, we aim to minimize the least squares error of the observed ratings (and regularize):\n",
    "\n",
    "$$\n",
    "\\min_{X,Y} \\sum_{r_{ui} \\text{observed}}(r_{ui} - x_{u}^{T}y_i)^2 + \\lambda \\left( \\sum_{u} \\|x_u\\|^2 + \\sum_{i} \\|y_i\\|^2 \\right) \n",
    "$$\n",
    "\n",
    "Notice that this objective is non-convex (because of the $x_{u}^{T} y_i$ term); in fact it’s NP-hard to optimize. Gradient descent can be used as an approximate approach here, however it turns out to be slow and costs lots of iterations. Note however, that if we fix the set of variables $X$ and treat them as constants, then the objective is a convex function of $Y$ and vice versa. Our approach will therefore be to fix $Y$ and optimize $X$, then fix $X$ and optimize $Y$, and repeat until convergence. This approach is known as **ALS (Alternating Least Squares)**. For our objective function, the alternating least squares algorithm can be expressed with this simple pseudo-code:"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "\n",
    "**Initialize** $X$, $Y$\n",
    "\n",
    "**while(convergence is not true) do**\n",
    "\n",
    "\n",
    "**for** $u = 1 \\cdots n$ **do**\n",
    "\n",
    "$x_u = \\left( \\sum_{r_ui \\in r_{u*}} y_i y_{i}^{T} + \\lambda I_k \\right)^{-1} \\sum_{r_ui \\in r_{u*}} r_{ui} y_i $ \n",
    "   \n",
    "**end for**\n",
    "\n",
    "**for** $u = 1 \\cdots n$ **do**\n",
    "\n",
    "$y_i = \\left( \\sum_{r_ui \\in r_{*i}} x_u x_{u}^{T} + \\lambda I_k \\right)^{-1} \\sum_{r_ui \\in r_{*i}} r_{ui} x_u $ \n",
    "   \n",
    "**end for**\n",
    "\n",
    "\n",
    "**end while**\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "For a single machine, we can analyze the computational cost of this algorithm. Updating each $x_u$ will cost $O(n_u k^2 + k^3)$, where $n_u$ is the number of items rated by user $u$, and similarly updating each $y_i$ will cost $O(n_i k^2 + k^3)$, where $n_i$ is the number of users that have rated item $i$.\n",
    "\n",
    "\n",
    "Once we’ve computed the matrices $X$ and $Y$, there are several ways compute a prediction. The first is to do what was discussed before, which is to simply predict $ r_{ui} \\approx x_{u}^{T}y_i$ for each user $u$ and item $i$. \n",
    "This approach will cost $O(nmk)$ if we’d like to estimate every user-item pair. \n",
    "\n",
    "However, this approach is prohibitively expensive for most real-world datasets. A second (and more holistic) approach is to use the $x_u$ and $y_i$ as features in another learning algorithm, incorporating these features with others that are relevant to the prediction task."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Further readings\n",
    "Other methods for matrix factorization include:\n",
    "\n",
    "* Low Rank Approximation and Regression in Input Sparsity Time, by Kenneth L. Clarkson, David P. Woodruff. http://arxiv.org/abs/1207.6365\n",
    "* Generalized Low Rank Models (GLRM), by Madeleine Udell, Corinne Horn, Reza Zadeh, Stephen Boyd. http://arxiv.org/abs/1410.0342\n",
    "* Matrix Completion and Low-Rank SVD via Fast Alternating Least Squares, by Trevor Hastie, Rahul Mazumder, Jason D. Lee, Reza Zadeh . Statistics Department and ICME, Stanford University, 2014. http://stanford.edu/~rezab/papers/fastals.pdf"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 3. Usecase : Music recommender system\n",
    "\n",
    "In this usecase, we use the data of users and artists in the previous sections to build a statistical model to recommend artists for users.\n",
    " \n",
    "## 3.1 Requirements\n",
    "According to the properties of data, we need to choose a recommender algorithm that is suitable for this implicit feedback data. It means that the algorithm should learn without access to user or artist attributes such as age, genre,.... Therefore, an algorithm of type `collaborative filtering` is the best choice.\n",
    "\n",
    "Second, in the data, there are some users that have listened to only 1 artist. We need an algorithm that might provide decent recommendations to even these users. After all, at some point, every user starts out with just one play at some point! \n",
    "\n",
    "Third, we need an algorithm that scales, both in its ability to build large models, and to create recommendations quickly.\n",
    "\n",
    "From these requirement, we can choose using ALS algorithm in [Implicit](https://implicit.readthedocs.io/en/latest/)."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### Question 6\n",
    "\n",
    "The library choosen for implementing ALS has a strict requirement in terms of data format. In particular `userID` and `artistID` should be both incremetally assigned starting from 0 (e.g. the first valid `artistID` is 90 and it has to correspond to 0). In these next questions we will try to fix this issue.\n",
    "\n",
    "#### Question 6.1\n",
    "\n",
    "<div class=\"alert alert-info\">\n",
    "Create a table with incremental ID for artists and print the first 15 entries.\n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 93,
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>originalArtistID</th>\n",
       "      <th>incrementalArtistID</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>24</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>46</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>121</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>151</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>227</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>342</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>566</td>\n",
       "      <td>6</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>619</td>\n",
       "      <td>7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>737</td>\n",
       "      <td>8</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>787</td>\n",
       "      <td>9</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>798</td>\n",
       "      <td>10</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>938</td>\n",
       "      <td>11</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>1208</td>\n",
       "      <td>12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>1404</td>\n",
       "      <td>13</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>1435</td>\n",
       "      <td>14</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    originalArtistID  incrementalArtistID\n",
       "0                 24                    0\n",
       "1                 46                    1\n",
       "2                121                    2\n",
       "3                151                    3\n",
       "4                227                    4\n",
       "5                342                    5\n",
       "6                566                    6\n",
       "7                619                    7\n",
       "8                737                    8\n",
       "9                787                    9\n",
       "10               798                   10\n",
       "11               938                   11\n",
       "12              1208                   12\n",
       "13              1404                   13\n",
       "14              1435                   14"
      ]
     },
     "execution_count": 93,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# create an ID mapping DF\n",
    "artistMappingDF = pd.DataFrame()\n",
    "\n",
    "# create a column with the original unique artist ID's\n",
    "artistMappingDF['originalArtistID']  = newUserArtistDF['realArtistID'].unique()\n",
    "\n",
    "# create a column with the incremental artist ID\n",
    "artistMappingDF['incrementalArtistID'] = [i for i in range(len(newUserArtistDF['realArtistID'].unique()))]\n",
    "\n",
    "artistMappingDF[:15]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "#### Question 6.2 \n",
    "\n",
    "<div class=\"alert alert-info\">\n",
    "Create a table with incremental ID for users and print the first 15 entries.\n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 94,
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>originalUserID</th>\n",
       "      <th>incrementalUserID</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>90</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>98</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>116</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>120</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>128</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>139</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>277</td>\n",
       "      <td>6</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>321</td>\n",
       "      <td>7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>340</td>\n",
       "      <td>8</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>350</td>\n",
       "      <td>9</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>384</td>\n",
       "      <td>10</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>519</td>\n",
       "      <td>11</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>521</td>\n",
       "      <td>12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>536</td>\n",
       "      <td>13</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>581</td>\n",
       "      <td>14</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    originalUserID  incrementalUserID\n",
       "0               90                  0\n",
       "1               98                  1\n",
       "2              116                  2\n",
       "3              120                  3\n",
       "4              128                  4\n",
       "5              139                  5\n",
       "6              277                  6\n",
       "7              321                  7\n",
       "8              340                  8\n",
       "9              350                  9\n",
       "10             384                 10\n",
       "11             519                 11\n",
       "12             521                 12\n",
       "13             536                 13\n",
       "14             581                 14"
      ]
     },
     "execution_count": 94,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# create an ID mapping DF\n",
    "userMappingDF = pd.DataFrame()\n",
    "\n",
    "# create a column with the original unique user ID's\n",
    "userMappingDF['originalUserID'] = newUserArtistDF['userID'].unique()\n",
    "\n",
    "# create a column with the incremental user ID\n",
    "userMappingDF['incrementalUserID'] = [0 + i for i in range(len(newUserArtistDF['userID'].unique()))]\n",
    "\n",
    "userMappingDF[:15]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "#### Question 6.3 \n",
    "\n",
    "<div class=\"alert alert-info\">\n",
    "Starting from the clean dataset that we already have, find a way to replace the original IDs with the new ones. Print the first 15 entries in this new table.\n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 95,
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>playCount</th>\n",
       "      <th>incrementalArtistID</th>\n",
       "      <th>incrementalUserID</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>4</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1</td>\n",
       "      <td>3</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>16</td>\n",
       "      <td>4</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>1</td>\n",
       "      <td>5</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>1</td>\n",
       "      <td>6</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>1</td>\n",
       "      <td>7</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>1</td>\n",
       "      <td>8</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>7</td>\n",
       "      <td>9</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>72</td>\n",
       "      <td>10</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>5</td>\n",
       "      <td>11</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>67</td>\n",
       "      <td>12</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>1</td>\n",
       "      <td>13</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>87</td>\n",
       "      <td>14</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    playCount  incrementalArtistID  incrementalUserID\n",
       "0           1                    0                  0\n",
       "1           4                    1                  0\n",
       "2           1                    2                  0\n",
       "3           1                    3                  0\n",
       "4          16                    4                  0\n",
       "5           1                    5                  0\n",
       "6           1                    6                  0\n",
       "7           1                    7                  0\n",
       "8           1                    8                  0\n",
       "9           7                    9                  0\n",
       "10         72                   10                  0\n",
       "11          5                   11                  0\n",
       "12         67                   12                  0\n",
       "13          1                   13                  0\n",
       "14         87                   14                  0"
      ]
     },
     "execution_count": 95,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Join the cleaned dataframe with the artist ID mapping dataframe over the original artist ID's\n",
    "rescaledUserArtistDF = pd.merge(newUserArtistDF, artistMappingDF, \n",
    "                                left_on='realArtistID', \n",
    "                                right_on='originalArtistID').drop(['realArtistID','originalArtistID'], axis=1)\n",
    "\n",
    "# Join the cleaned dataframe with the user ID mapping dataframe over the original user ID's\n",
    "rescaledUserArtistDF = pd.merge(rescaledUserArtistDF, userMappingDF, \n",
    "                                left_on='userID',\n",
    "                                right_on='originalUserID').drop(['userID','originalUserID'], axis=1)\n",
    "\n",
    "rescaledUserArtistDF[:15]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now, we don't need no more `userArtistDF` and `newUserArtistDF`. We can delete them and free some memory"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "del userArtistDF\n",
    "del newUserArtistDF"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "## 3.4 Training our statistical model\n",
    "To train a model using ALS, we must use a sparse matrix as an input. Implicit uses the class sparse matrix to support the construction of a distributed preference matrix."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Question 7\n",
    "\n",
    "The implicit library expects data as a item-user sparse matrix. Now we create two matricies, one for fitting the model (item-user) and one for recommendations (the transpose of item-user -- user-item)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 96,
   "metadata": {
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "outputs": [],
   "source": [
    "import scipy.sparse as sparse\n",
    "import implicit\n",
    "\n",
    "def to_sparse(userArtistDF):\n",
    "    playCount = userArtistDF['playCount'].astype(float)\n",
    "    artistID = userArtistDF['incrementalArtistID']\n",
    "    userID = userArtistDF['incrementalUserID']\n",
    "    sparse_item_user = sparse.csr_matrix((playCount, (artistID, userID)))\n",
    "\n",
    "    return sparse_item_user\n",
    "\n",
    "sparse_item_user = to_sparse(rescaledUserArtistDF)\n",
    "sparse_user_item = sparse_item_user.T"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "#### Question 7.1\n",
    "<div class=\"alert alert-info\">\n",
    "Train a model trained by using `implicit.als.AlternatingLeastSquares()`.\n",
    "</div>\n",
    "\n",
    "Read **carefully** the documentation of this class."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We can also use some additional parameters to adjust the quality of the model. For now, let's set \n",
    "\n",
    "- `factors = 10`\n",
    "- `iterations = 5`\n",
    "- `regularization = 0.01`\n",
    "- `alpha = 1.0`\n",
    "\n",
    "to build the model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 97,
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [00:35<00:00,  8.05s/it]\n"
     ]
    }
   ],
   "source": [
    "# Calculate the confidence matrix by multiplying the item/user matrix by our alpha value.\n",
    "data_conf = (sparse_item_user * 1.0).astype('double')\n",
    "\n",
    "# Create the model according to the given parameters\n",
    "model = implicit.als.AlternatingLeastSquares(factors=10, regularization=0.01, iterations=5)\n",
    "\n",
    "# Fit the model\n",
    "model.fit(data_conf, show_progress = True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "#### Question 7.5\n",
    "<div class=\"alert alert-info\">\n",
    "Print the user features for the first user\n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 98,
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([-0.00682594,  0.49205855,  0.03019511,  0.01725622, -0.31710687,\n",
       "       -0.15663357,  0.10828613, -0.20233256,  0.24350643,  0.1440268 ],\n",
       "      dtype=float32)"
      ]
     },
     "execution_count": 98,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# display the k = 10 latent features for the first user\n",
    "model.user_factors[0]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### Question 8\n",
    "<div class=\"alert alert-info\">\n",
    "Show the top-5 artist names recommendated for a given user, for example: `2093760` (please, try also with different users!).\n",
    "</div>\n",
    "\n",
    "**HINT**: The recommendations can be given by function `recommend()`. These recommendations are only artist ids. You have to map them to artist names by using data in `artist_data.txt`. Remember that we have transformed our IDs, so what you need to do is `userID -> incrementalUserID -> recommend() -> incrementalArtistID -> artistID -> name`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 99,
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [],
   "source": [
    "# Retrieve the incrementalUserID from a userID\n",
    "def get_incrementalUserID(userID):\n",
    "    return userMappingDF[userMappingDF.originalUserID == userID]['incrementalUserID'].iloc[0]\n",
    "\n",
    "# Retrieve the original userID from an incrementalUserID\n",
    "def get_userID(incrementalUserID):\n",
    "    return userMappingDF[userMappingDF.incrementalUserID == incrementalUserID]['originalUserID'].iloc[0]\n",
    "\n",
    "# Retrieve the incrementalArtistID from an artistID\n",
    "def get_incrementalArtistID(artisID):\n",
    "    return artistMappingDF[artistMappingDF.originalArtistID == artisID]['incrementalArtistID'].iloc[0]\n",
    "\n",
    "# Retrieve the original artistID from an incrementalArtistID\n",
    "def get_artistID(incrementalArtistID):\n",
    "    return artistMappingDF[artistMappingDF.incrementalArtistID == incrementalArtistID]['originalArtistID'].iloc[0]\n",
    "\n",
    "# Retrive the artist name\n",
    "def get_artist_name(artistID):\n",
    "    return artistDF[artistDF.artistID == artistID]['name'].iloc[0]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Testing Recommendations for Different Types of Users"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 107,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "User  2093760\n",
      "Play Count: 14\n",
      "Top 1\n",
      "Artist Name: 50 Cent\n",
      "Score: 0.022041637\n",
      "===========================\n",
      "Top 2\n",
      "Artist Name: 2Pac\n",
      "Score: 0.021421568\n",
      "===========================\n",
      "Top 3\n",
      "Artist Name: Snoop Dogg\n",
      "Score: 0.021306906\n",
      "===========================\n",
      "Top 4\n",
      "Artist Name: Dr. Dre\n",
      "Score: 0.0208239\n",
      "===========================\n",
      "Top 5\n",
      "Artist Name: Jay-Z\n",
      "Score: 0.02076098\n",
      "===========================\n",
      "===========================\n",
      "\n",
      "User  1040331 -> Top play count user\n",
      "Top 1\n",
      "Artist Name: Jurassic 5\n",
      "Score: 1.3356354\n",
      "===========================\n",
      "Top 2\n",
      "Artist Name: DJ Danger Mouse\n",
      "Score: 1.3007516\n",
      "===========================\n",
      "Top 3\n",
      "Artist Name: Blind Melon\n",
      "Score: 1.2764741\n",
      "===========================\n",
      "Top 4\n",
      "Artist Name: Mitch Hedberg\n",
      "Score: 1.2702206\n",
      "===========================\n",
      "Top 5\n",
      "Artist Name: Soul Coughing\n",
      "Score: 1.2673836\n",
      "===========================\n",
      "===========================\n",
      "\n",
      "User  2443548 -> Bottom play count user\n",
      "1\n",
      "Top 1\n",
      "Artist Name: Mae\n",
      "Score: 0.0031595451\n",
      "===========================\n",
      "Top 2\n",
      "Artist Name: The Starting Line\n",
      "Score: 0.0030924124\n",
      "===========================\n",
      "Top 3\n",
      "Artist Name: Copeland\n",
      "Score: 0.0030714208\n",
      "===========================\n",
      "Top 4\n",
      "Artist Name: Fall Out Boy\n",
      "Score: 0.0030710357\n",
      "===========================\n",
      "Top 5\n",
      "Artist Name: The Academy Is...\n",
      "Score: 0.0030355891\n",
      "===========================\n",
      "===========================\n",
      "\n",
      "User  90 -> Top curious user\n",
      "Top 1\n",
      "Artist Name: Miles Davis\n",
      "Score: 0.37792236\n",
      "===========================\n",
      "Top 2\n",
      "Artist Name: Amon Tobin\n",
      "Score: 0.37310824\n",
      "===========================\n",
      "Top 3\n",
      "Artist Name: Thievery Corporation\n",
      "Score: 0.37303177\n",
      "===========================\n",
      "Top 4\n",
      "Artist Name: Prefuse 73\n",
      "Score: 0.36196524\n",
      "===========================\n",
      "Top 5\n",
      "Artist Name: RJD2\n",
      "Score: 0.36110282\n",
      "===========================\n",
      "===========================\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Recommend artists for user with ID 2093760\n",
    "user_id = 2093760\n",
    "top = 1\n",
    "print(\"User \",2093760)\n",
    "print(\"Play Count:\", rankedUsers[rankedUsers.userID == 2093760]['playCount'].values[0])\n",
    "for incremental_artist_id, score in model.recommend(userid = get_incrementalUserID(user_id), user_items = sparse_user_item, N =5):\n",
    "    print(\"Top\",top)\n",
    "    top += 1\n",
    "    print(\"Artist Name:\",get_artist_name(get_artistID(incremental_artist_id)))\n",
    "    print(\"Score:\",score)\n",
    "    print(\"===========================\")\n",
    "print(\"===========================\\n\")\n",
    "\n",
    "# Recommend artists for user with highest play count\n",
    "user_id = userPlayCount.iloc[0]['userID']\n",
    "top = 1\n",
    "print(\"User \", user_id, \"-> Top play count user\")\n",
    "for incremental_artist_id, score in model.recommend(userid = get_incrementalUserID(user_id), user_items = sparse_user_item, N =5):\n",
    "    print(\"Top\",top)\n",
    "    top += 1\n",
    "    print(\"Artist Name:\",get_artist_name(get_artistID(incremental_artist_id)))\n",
    "    print(\"Score:\",score)\n",
    "    print(\"===========================\")\n",
    "print(\"===========================\\n\")\n",
    "\n",
    "# Recommend artists for user with lowest play count\n",
    "user_id = rankedUsers.iloc[-1]['userID']\n",
    "top = 1\n",
    "print(\"User \", user_id, \"-> Bottom play count user\")\n",
    "print(rankedUsers.iloc[-1]['playCount'])\n",
    "for incremental_artist_id, score in model.recommend(userid = get_incrementalUserID(user_id), user_items = sparse_user_item, N =5):\n",
    "    print(\"Top\",top)\n",
    "    top += 1\n",
    "    print(\"Artist Name:\",get_artist_name(get_artistID(incremental_artist_id)))\n",
    "    print(\"Score:\",score)\n",
    "    print(\"===========================\")\n",
    "print(\"===========================\\n\")\n",
    " \n",
    "# Recommend artists for user with highest curiosity\n",
    "user_id = userCuriosity.iloc[0]['userID']\n",
    "top = 1\n",
    "print(\"User \",user_id, \"-> Top curious user\")\n",
    "for incremental_artist_id, score in model.recommend(userid = get_incrementalUserID(user_id), user_items = sparse_user_item, N =5):\n",
    "    print(\"Top\",top)\n",
    "    top += 1\n",
    "    print(\"Artist Name:\",get_artist_name(get_artistID(incremental_artist_id)))\n",
    "    print(\"Score:\",score)\n",
    "    print(\"===========================\")\n",
    "print(\"===========================\\n\")\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<div class=\"alert alert-success\" style=\"background-color: AliceBlue; border-color: CornflowerBlue \">\n",
    "Above, we have used our trained model to recommend artists for different kinds of users in order to observe the different recommendations along with the confidence of these recommendations.<br>\n",
    "* With the user having an ID <b>2093760</b> and a play count <b>14</b>, we can see that the confidence of the recommendations are just around <b>0.02</b>.<br>\n",
    "* With the user having the <b>top play count</b> in our data, we can see that the confidence of the recommendations can reach approximately <b>22</b>. Thus we can start to see the impact of the <b>implicit feedback</b> on the recommendation confidence and quality.<br>\n",
    "* With the user having the <b>least play count</b> in our data, we can see that the confidence of the recommendations has significantly dropped to <b>0.002</b>. Thus can be explained by the fact that this user has not given sufficient feedback to the system to obtain good recommendations.<br>\n",
    "* Finnaly we have tested the recommendations with the <b>most curious user</b> in our dataset, and we can see that the confidence obtained was <b>1.5</b> which is lower that the that of the users with the highest play count but way higher than users with small feedback to the system.\n",
    "</div>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Extra: Plot the Relationships between the Recommendation Confidence and each of the User Curiosity and Play Count"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 109,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "147960"
      ]
     },
     "execution_count": 109,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "userPlayCount.count().values[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 110,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x864 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Create tuples (number of play counts, confidence of the 1st recomendation) for 149 users ranging from top to bottom play count\n",
    "'''\n",
    "The top ranked users were removed\n",
    "for the sake of making the plot more clear\n",
    "as their confidence ruins the confidence scale\n",
    "but such users' recommendations with their confidence\n",
    "can be viewed in the previous part\n",
    "'''\n",
    "playCount_Confidence = [(rankedUsers.iloc[i]['playCount'], model.recommend(userid = get_incrementalUserID(userPlayCount.iloc[i]['userID']), user_items = sparse_user_item, N =1)[0][1])  for i in range(0, userPlayCount.count().values[0]-1, 1000)]\n",
    "\n",
    "\n",
    "# Plot PlayCounts and accuracy of the recommendation per user\n",
    "\n",
    "fig = plt.figure(1)\n",
    "\n",
    "\n",
    "ax1 = plt.subplot(2, 1, 1)\n",
    "\n",
    "ax1.semilogx()\n",
    "plt.scatter(*zip(*playCount_Confidence))\n",
    "ax1.set_xlabel('Number of Plays')\n",
    "ax1.set_ylabel('Recommendation Confidence')\n",
    "ax1.set_title('Relation between Play Count and Confidence in the Recommendation per User')\n",
    "\n",
    "\n",
    "#Create tuples (number of diferent artist the user has search, confidence of the 1st recomendation) for the 149 users ranging from top to bottom curiosity\n",
    "curiosity_Confidence = [(rankedUsersCuriosity.iloc[i]['realArtistID'], model.recommend(userid = get_incrementalUserID(rankedUsersCuriosity.iloc[i]['userID']), user_items = sparse_user_item, N =1)[0][1])  for i in range(0, userPlayCount.count().values[0]-1, 1000)]\n",
    "\n",
    "\n",
    "\n",
    "ax2 = plt.subplot(2, 1, 2)\n",
    "\n",
    "ax2.semilogx()\n",
    "plt.scatter(*zip(*curiosity_Confidence))\n",
    "ax2.set_xlabel('Number of Artists a User has Searched')\n",
    "ax2.set_ylabel('Recommendation Confidence')\n",
    "ax2.set_title('Relation between the Curiosity and Confidence in the Recommendation per User')\n",
    "\n",
    "fig.set_size_inches(w=12,h=12)\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<div class=\"alert alert-success\" style=\"background-color: AliceBlue; border-color: CornflowerBlue \"> \n",
    "As we can see in the <b>first plot above</b>, there is a relation between the play counts of users, and the score of the recommendations offered by the model (<b>confidence</b>). If we have know more information about this user, we can recommend something better to him/her, a higher play count implies a stronger interaction with the system, thus, more information and feedback, and as a result, more confident recommendations.\n",
    "    \n",
    "In the <b>second plot</b> where we visualize the recommendation confidence as a function of the curiosity of the user. We can also see that confidence increases with increasing curiosity. Therefore, we can say that also if the user is more curious, the recommendations will be better.\n",
    "</div>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "## 3.5 Evaluating Recommendation Quality \n",
    "\n",
    "In this section, we study how to evaluate the quality of our model. It's hard to say how good the recommendations are.\n",
    "One of serveral methods approach to evaluate  a recommender based on its ability to rank good items (artists) high in a list of recommendations. The problem is how to define \"good artists\". Currently, by training all data, \"good artists\" is defined as \"artists the user has listened to\", and the recommender system has already received all of this information as input. It could trivially return the users previously-listened artists as top recommendations and score perfectly. Indeed, this is not useful, because the recommender's is used to recommend artists that the user has **never** listened to. \n",
    "\n",
    "To overcome that problem, we can hide the some of the artist play data and only use the rest to train model. Then, this held-out data can be interpreted as a collection of \"good\" recommendations for each user. The recommender is asked to rank all items in the model, and the rank of the held-out artists are examined. Ideally the recommender places all of them at or near the top of the list.\n",
    "\n",
    "The recommender's score can then be computed by comparing all held-out artists' ranks to the rest.  The fraction of pairs where the held-out artist is ranked higher is its score. 1.0 is perfect, 0.0 is the worst possible score, and 0.5 is the expected value achieved from randomly ranking artists. \n",
    "\n",
    "AUC(Area Under the Curve) can be used as a metric to evaluate model. It is also viewed as the probability that a randomly-chosen \"good\" artist ranks above a randomly-chosen \"bad\" artist.\n",
    "\n",
    "Next, we split the training data into 2 parts: `trainData` and `cvData` with ratio 0.9:0.1 respectively, where `trainData` is the dataset that will be used to train model. Then we write a function to calculate AUC to evaluate the quality of our model.\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Question 9\n",
    "\n",
    "#### Question 9.1\n",
    "<div class=\"alert alert-info\">\n",
    "Split the data into `trainData` and `cvData` with ratio 0.9:0.1 and use the first part to train a statistic model with:\n",
    "<ul>\n",
    "<li>`rank`=10</li>\n",
    "<li>`iterations`=5</li>\n",
    "<li>`lambda_`=0.01</li>\n",
    "<li>`alpha`=1.0</li>\n",
    "</ul>\n",
    "</div>\n",
    "\n",
    "***Hint***: have a look to `train_test_split()` in `sklearn.model_selection`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 113,
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.5/dist-packages/sklearn/model_selection/_split.py:2026: FutureWarning: From version 0.21, test_size will always complement train_size unless both are specified.\n",
      "  FutureWarning)\n",
      "100%|██████████| 5.0/5 [00:33<00:00,  7.44s/it]\n"
     ]
    }
   ],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "# Create train/test split 0.9:0.1\n",
    "trainUserArtistDF, testUserArtistDF = train_test_split(rescaledUserArtistDF, train_size = 0.9)\n",
    "\n",
    "# Transform to sparse matrix\n",
    "train_sparse_item_user = to_sparse(trainUserArtistDF)\n",
    "train_sparse_user_item = train_sparse_item_user.T\n",
    "\n",
    "# Compute the confidence matrix\n",
    "data_conf = (train_sparse_item_user * 1.0 ).astype('double')\n",
    "\n",
    "# Create the model\n",
    "model = implicit.als.AlternatingLeastSquares(factors=10, regularization=0.01, iterations=5)\n",
    "\n",
    "# Train\n",
    "model.fit(data_conf, show_progress = True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Area under the ROC curve: a function to compute it"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 114,
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [],
   "source": [
    "from multiprocessing import Pool\n",
    "import warnings\n",
    "from sklearn.metrics import roc_auc_score\n",
    "import random\n",
    "from tqdm import tqdm\n",
    "\n",
    "\n",
    "def findNumElementsSmallerThan(arr, x, start=0):\n",
    "        left = start\n",
    "        right = len(arr) -1\n",
    "        # if x is bigger than the biggest element in arr\n",
    "        if start > right or x > arr[right]:\n",
    "            return right + 1\n",
    "        mid = -1\n",
    "        while left <= right:\n",
    "            mid = (left + right) // 2\n",
    "            if arr[mid] < x:\n",
    "                left = mid + 1\n",
    "            elif arr[mid] > x:\n",
    "                right = mid - 1\n",
    "            else:\n",
    "                while mid-1 >= start and arr[mid-1] == x:\n",
    "                    mid -= 1\n",
    "                return mid\n",
    "        return mid if arr[mid] > x else mid + 1\n",
    "\n",
    "def _compute_auc(incremental_user_id):\n",
    "    positive_products = testUserArtistDF[testUserArtistDF['incrementalUserID'] == incremental_user_id]['incrementalArtistID'].values\n",
    "    positive_ranking = model.rank_items(incremental_user_id, \n",
    "                                        train_sparse_user_item, \n",
    "                                        positive_products)\n",
    "    positive_ranking = list(map(lambda x: x[1], positive_ranking))\n",
    "    if len(positive_products) <= 1: return \n",
    "        \n",
    "    train_data_user = trainUserArtistDF[trainUserArtistDF['incrementalUserID'] == incremental_user_id]['incrementalArtistID'].values\n",
    "    \n",
    "    if len(train_data_user) < 1: return\n",
    "    negative_products = np.random.choice(train_data_user, len(positive_products))  \n",
    "    if len(negative_products) <= 1: return\n",
    "        \n",
    "    if len(negative_products) != len(positive_products): return \n",
    "    \n",
    "    negative_ranking = model.rank_items(incremental_user_id, \n",
    "                                        train_sparse_user_item, \n",
    "                                        negative_products)\n",
    "    \n",
    "    negative_ranking = list(map(lambda x: x[1], negative_ranking))\n",
    "   \n",
    "    correct = 0\n",
    "    total = len(positive_ranking)*len(negative_ranking)\n",
    "\n",
    "    for positive in positive_ranking:\n",
    "        correct += findNumElementsSmallerThan(negative_ranking, positive)\n",
    "    return float(correct) / total\n",
    "\n",
    "def compute_auc(test_users_ratio=1, parallelism=4):  \n",
    "    \"\"\"\n",
    "        Computes the AUC of a train/test split\n",
    "        Note:\n",
    "            It assumes `trainUserArtistDF`, `testUserArtistDf` and `model` to be defined \n",
    "            globally\n",
    "        Arguments:\n",
    "            test_users_ratio (float): Ratio of users to be tested (0, 1]\n",
    "    \"\"\"\n",
    "    if test_users_ratio <=0 or test_users_ratio > 1:\n",
    "        warnings.warn('Parameter test_users_ratio has to be in (0, 1]. Setting test_users_ratio=1...')\n",
    "        test_users_ratio = 1\n",
    "    \n",
    "    pool = Pool(parallelism)\n",
    "    roc_auc = []\n",
    "    list_of_users = testUserArtistDF['incrementalUserID'].unique()[::int(1/test_users_ratio)]\n",
    "    for res in tqdm(pool.imap_unordered(_compute_auc, list_of_users, 64), total=len(list_of_users)):\n",
    "        roc_auc.append(res)\n",
    "    \n",
    "    pool.close()\n",
    "    pool.join()\n",
    "\n",
    "    return np.array(list(filter(None.__ne__, roc_auc)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "####  Question 9.2\n",
    "<div class=\"alert alert-info\">\n",
    "Using `compute_auc()`, compute the AUC of the trained model.\n",
    "</div>\n",
    "\n",
    "***Note***: `compute_auc()` is computational expensive. To test if everything works, try to first compute the AUC for a small subset of users (10%) and than move to the entire set."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 134535/134535 [21:54<00:00, 102.35it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC = 0.8677\n"
     ]
    }
   ],
   "source": [
    "roc_auc = compute_auc(test_users_ratio = 1)\n",
    "print('Average AUC = %.4f' % roc_auc.mean())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "#### Question 9.3 \n",
    "\n",
    "The average AUC doesn't give the full picture here. Try to plot the distribution and the cumulative distribution of the AUC for all the users. Can you comment on these plots? "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x432 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, (ax0, ax1) = plt.subplots(2,1)\n",
    "\n",
    "fig.set_size_inches(12,6)\n",
    "colors = {0:\"red\", 1: \"orange\", 2:\"yellow\", 3:\"darkgreen\", 4:\"cyan\"}\n",
    "\n",
    "# Plot distribution\n",
    "ax0.hist(roc_auc, bins='auto', density=True)\n",
    "\n",
    "# Plot cumulative distribution\n",
    "ax1.hist(roc_auc, bins='auto', density=True, cumulative=True)\n",
    "\n",
    "# Plot a vertical line on the average (distribution plot)\n",
    "ax0.axvline(roc_auc.mean(), label='Average AUC = %.3f' % (roc_auc.mean()*100), color = 'yellow')\n",
    "\n",
    "# Plot some percentile on the cumulative distribution\n",
    "for i, quantile in enumerate([.05, 0.25, .5, .75, .95]):\n",
    "    ax1.axvline(np.percentile(roc_auc, quantile*100), label='%.0fth perc.' % (quantile*100), color = colors[i])\n",
    "    ax1.axhline(quantile, linestyle='-')\n",
    "    \n",
    "ax0.grid(True)\n",
    "ax0.set_title('Distribution of AUC for all the users')\n",
    "ax1.set_title('Cumulative distribution of AUC for all the users')\n",
    "ax1.annotate('MEDIAN', (0.2,0.5), size = 15)     \n",
    "ax0.legend()\n",
    "ax1.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<div class=\"alert alert-success\" style=\"background-color: AliceBlue; border-color: CornflowerBlue \"> \n",
    "Above, we obtain 2 plots for the the Area Under the Curve distribution for our test data. The <b>first plot</b> shows that a high number of users (aroud 58 users) in our test data actually have received a <b>100% AUC</b>. This result actually shows good potential for our recommender system as it is providing good rankings for the artist of the test data for the corresponding uers in the test data. Other values to point out for AUC are for example 50% and 67% which ranges over about 6 users. But also it is apparent values of AUC for our test data are somehow <b>skewed to the right</b> towards the higher values of AUC and this is also signified by the value of the <b>mean</b> which is <b>86.77%</b>.<br><br>\n",
    "\n",
    "In the <b>second plot</b> we can see the AUC distribution in a <b>Comulative</b> fashion. We've pointed out <b>5</b> percentiles (5%, 25%, 50%, 75%, 95%) as show in the plot above and we can extract the following observations:<br>\n",
    "* only 10% of AUC values are 50% or less.<br>\n",
    "* 25% of AUC values are 82% or less.<br>\n",
    "* 50% of AUC values are 93% or less. (<b>Median</b>).<br>\n",
    "* 75% of AUC values are at 100% or less which means that also all the following percentiles are at this AUC value or less.<br>\n",
    "Thus, also from the comulative distribution of the area under the curve for our test data, we can see that our recommender system is quite performing well as more than half the users in our test data have more than 90% pair-proportion where the ranked artists are ranking higher than the other artists in the training set.<br>\n",
    "<b> BUT CAN WE IMPROVE THE PERFORMANCE EVEN FURTHER?</b>\n",
    "</div>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "## 3.6 Personalized recommendations with ALS\n",
    "\n",
    "In the previous section, we build our models with some given paramters without any knowledge about them. Actually, choosing the best parameters' values is very important. It can significantly affect the quality of models. Especially, with the current implementation of ALS in MLLIB, these parameters are not learned by the algorithm, and must be chosen by the caller. The following parameters should get consideration before training models:\n",
    "\n",
    "* `rank = 10`: the number of latent factors in the model, or equivalently, the number of columns $k$ in the user-feature and product-feature matrices. In non-trivial cases, this is also their rank. \n",
    "\n",
    "* `iterations = 5`: the number of iterations that the factorization runs. Instead of runing the algorithm until RMSE converged which actually takes very long time to finish with large datasets, we only let it run in a given number of iterations. More iterations take more time but may produce a better factorization.\n",
    "\n",
    "* `lambda_ = 0.01`: a standard overfitting parameter. Higher values resist overfitting, but values that are too high hurt the factorization's accuracy.\n",
    "\n",
    "*  `alpha = 1.0`: controls the relative weight of observed versus unobserved userproduct interactions in the factorization. \n",
    "\n",
    "Although all of them have impact on the models' quality, `iterations` is more of a constraint on resources used in the factorization. So, `rank`, `lambda_` and `alpha` can be considered hyperparameters to the model. \n",
    "We will try to find \"good\" values for them. Indeed, the values of hyperparameter are not necessarily optimal. Choosing good hyperparameter values is a common problem in machine learning. The most basic way to choose values is to simply try combinations of values and evaluate a metric for each of them, and choose the combination that produces the best value of the metric. \n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### Question 10\n",
    "\n",
    "#### Question 10.1\n",
    "<div class=\"alert alert-info\">\n",
    "For simplicity, assume that we want to explore the following parameter space: $ rank \\in \\{10, 50\\}$, $lambda\\_ \\in \\{1.0, 0.0001\\}$ and $alpha \\in \\{1.0, 40.0\\}$.\n",
    "\n",
    "Find the best combination of them in terms of the highest AUC value.\n",
    "</div>\n",
    "\n",
    "***Hint***: have a look to `ParameterGrid()` in `sklearn.model_selection` "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "scrolled": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Combination 1\n",
      "==========================================\n",
      "Factors = 10\n",
      "Regularization = 1\n",
      "Alpha = 1\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [01:08<00:00, 15.28s/it]\n",
      "100%|██████████| 13460/13460 [02:08<00:00, 104.59it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.8679575985359057\n",
      "==========================================\n",
      "Combination 2\n",
      "==========================================\n",
      "Factors = 10\n",
      "Regularization = 0.01\n",
      "Alpha = 1\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [01:06<00:00, 14.70s/it]\n",
      "100%|██████████| 13460/13460 [02:13<00:00, 101.08it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.8675695595504007\n",
      "==========================================\n",
      "Combination 3\n",
      "==========================================\n",
      "Factors = 10\n",
      "Regularization = 0.0001\n",
      "Alpha = 1\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [01:09<00:00, 15.33s/it]\n",
      "100%|██████████| 13460/13460 [02:09<00:00, 104.32it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.8676492875697928\n",
      "==========================================\n",
      "Combination 4\n",
      "==========================================\n",
      "Factors = 25\n",
      "Regularization = 1\n",
      "Alpha = 1\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [02:22<00:00, 31.85s/it]\n",
      "100%|██████████| 13460/13460 [02:13<00:00, 100.62it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.8682167959572555\n",
      "==========================================\n",
      "Combination 5\n",
      "==========================================\n",
      "Factors = 25\n",
      "Regularization = 0.01\n",
      "Alpha = 1\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [02:16<00:00, 29.93s/it]\n",
      "100%|██████████| 13460/13460 [02:13<00:00, 100.81it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.8665391246995592\n",
      "==========================================\n",
      "Combination 6\n",
      "==========================================\n",
      "Factors = 25\n",
      "Regularization = 0.0001\n",
      "Alpha = 1\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [02:22<00:00, 31.73s/it]\n",
      "100%|██████████| 13460/13460 [02:13<00:00, 100.74it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.8700820419307032\n",
      "==========================================\n",
      "Combination 7\n",
      "==========================================\n",
      "Factors = 50\n",
      "Regularization = 1\n",
      "Alpha = 1\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [03:27<00:00, 44.86s/it]\n",
      "100%|██████████| 13460/13460 [02:14<00:00, 100.27it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.8594690601135792\n",
      "==========================================\n",
      "Combination 8\n",
      "==========================================\n",
      "Factors = 50\n",
      "Regularization = 0.01\n",
      "Alpha = 1\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [03:31<00:00, 46.50s/it]\n",
      "100%|██████████| 13460/13460 [02:16<00:00, 98.61it/s] \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.8603935501458251\n",
      "==========================================\n",
      "Combination 9\n",
      "==========================================\n",
      "Factors = 50\n",
      "Regularization = 0.0001\n",
      "Alpha = 1\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [03:50<00:00, 50.36s/it]\n",
      "100%|██████████| 13460/13460 [02:17<00:00, 97.67it/s] \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.8613427420547696\n",
      "==========================================\n",
      "Combination 10\n",
      "==========================================\n",
      "Factors = 10\n",
      "Regularization = 1\n",
      "Alpha = 15\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [01:07<00:00, 14.83s/it]\n",
      "100%|██████████| 13460/13460 [02:04<00:00, 107.83it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.8670100017144682\n",
      "==========================================\n",
      "Combination 11\n",
      "==========================================\n",
      "Factors = 10\n",
      "Regularization = 0.01\n",
      "Alpha = 15\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [01:09<00:00, 15.05s/it]\n",
      "100%|██████████| 13460/13460 [02:07<00:00, 105.55it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.8660394282536671\n",
      "==========================================\n",
      "Combination 12\n",
      "==========================================\n",
      "Factors = 10\n",
      "Regularization = 0.0001\n",
      "Alpha = 15\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [01:08<00:00, 15.28s/it]\n",
      "100%|██████████| 13460/13460 [02:13<00:00, 100.98it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.8651136206186315\n",
      "==========================================\n",
      "Combination 13\n",
      "==========================================\n",
      "Factors = 25\n",
      "Regularization = 1\n",
      "Alpha = 15\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [02:27<00:00, 32.22s/it]\n",
      "100%|██████████| 13460/13460 [02:13<00:00, 100.47it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.8591587643592021\n",
      "==========================================\n",
      "Combination 14\n",
      "==========================================\n",
      "Factors = 25\n",
      "Regularization = 0.01\n",
      "Alpha = 15\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [02:25<00:00, 32.27s/it]\n",
      "100%|██████████| 13460/13460 [02:09<00:00, 103.61it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.8614165440549005\n",
      "==========================================\n",
      "Combination 15\n",
      "==========================================\n",
      "Factors = 25\n",
      "Regularization = 0.0001\n",
      "Alpha = 15\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [02:25<00:00, 31.81s/it]\n",
      "100%|██████████| 13460/13460 [02:13<00:00, 100.72it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.8589305193053804\n",
      "==========================================\n",
      "Combination 16\n",
      "==========================================\n",
      "Factors = 50\n",
      "Regularization = 1\n",
      "Alpha = 15\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [03:43<00:00, 48.28s/it]\n",
      "100%|██████████| 13460/13460 [02:14<00:00, 100.24it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.8494841810372491\n",
      "==========================================\n",
      "Combination 17\n",
      "==========================================\n",
      "Factors = 50\n",
      "Regularization = 0.01\n",
      "Alpha = 15\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [03:41<00:00, 48.62s/it]\n",
      "100%|██████████| 13460/13460 [02:16<00:00, 98.54it/s] \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.8482920178801152\n",
      "==========================================\n",
      "Combination 18\n",
      "==========================================\n",
      "Factors = 50\n",
      "Regularization = 0.0001\n",
      "Alpha = 15\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [03:43<00:00, 49.13s/it]\n",
      "100%|██████████| 13460/13460 [02:13<00:00, 100.55it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.8459094452272459\n",
      "==========================================\n",
      "Combination 19\n",
      "==========================================\n",
      "Factors = 10\n",
      "Regularization = 1\n",
      "Alpha = 30\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [01:07<00:00, 14.93s/it]\n",
      "100%|██████████| 13460/13460 [02:10<00:00, 103.00it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.8635626771941356\n",
      "==========================================\n",
      "Combination 20\n",
      "==========================================\n",
      "Factors = 10\n",
      "Regularization = 0.01\n",
      "Alpha = 30\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [01:09<00:00, 15.18s/it]\n",
      "100%|██████████| 13460/13460 [02:13<00:00, 101.02it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.8650352261288408\n",
      "==========================================\n",
      "Combination 21\n",
      "==========================================\n",
      "Factors = 10\n",
      "Regularization = 0.0001\n",
      "Alpha = 30\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [01:09<00:00, 15.08s/it]\n",
      "100%|██████████| 13460/13460 [02:11<00:00, 102.71it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.864138731558585\n",
      "==========================================\n",
      "Combination 22\n",
      "==========================================\n",
      "Factors = 25\n",
      "Regularization = 1\n",
      "Alpha = 30\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [02:26<00:00, 32.43s/it]\n",
      "100%|██████████| 13460/13460 [02:16<00:00, 98.53it/s] \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.8558336201581241\n",
      "==========================================\n",
      "Combination 23\n",
      "==========================================\n",
      "Factors = 25\n",
      "Regularization = 0.01\n",
      "Alpha = 30\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [02:27<00:00, 32.13s/it]\n",
      "100%|██████████| 13460/13460 [02:06<00:00, 106.39it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.8565223269989986\n",
      "==========================================\n",
      "Combination 24\n",
      "==========================================\n",
      "Factors = 25\n",
      "Regularization = 0.0001\n",
      "Alpha = 30\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [02:28<00:00, 32.89s/it]\n",
      "100%|██████████| 13460/13460 [02:14<00:00, 100.04it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.8562263225637959\n",
      "==========================================\n",
      "Combination 25\n",
      "==========================================\n",
      "Factors = 50\n",
      "Regularization = 1\n",
      "Alpha = 30\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [03:46<00:00, 49.26s/it]\n",
      "100%|██████████| 13460/13460 [02:13<00:00, 100.50it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.8454478508300567\n",
      "==========================================\n",
      "Combination 26\n",
      "==========================================\n",
      "Factors = 50\n",
      "Regularization = 0.01\n",
      "Alpha = 30\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [03:30<00:00, 47.70s/it]\n",
      "100%|██████████| 13460/13460 [02:12<00:00, 101.33it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.8445484662988277\n",
      "==========================================\n",
      "Combination 27\n",
      "==========================================\n",
      "Factors = 50\n",
      "Regularization = 0.0001\n",
      "Alpha = 30\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [03:25<00:00, 45.49s/it]\n",
      "100%|██████████| 13460/13460 [02:14<00:00, 99.87it/s] \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.846365914172028\n",
      "==========================================\n",
      "Combination 28\n",
      "==========================================\n",
      "Factors = 10\n",
      "Regularization = 1\n",
      "Alpha = 40\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [01:09<00:00, 15.46s/it]\n",
      "100%|██████████| 13460/13460 [02:12<00:00, 101.88it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.8647323884567009\n",
      "==========================================\n",
      "Combination 29\n",
      "==========================================\n",
      "Factors = 10\n",
      "Regularization = 0.01\n",
      "Alpha = 40\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [01:08<00:00, 15.03s/it]\n",
      "100%|██████████| 13460/13460 [02:10<00:00, 102.93it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.8638726688495644\n",
      "==========================================\n",
      "Combination 30\n",
      "==========================================\n",
      "Factors = 10\n",
      "Regularization = 0.0001\n",
      "Alpha = 40\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [01:08<00:00, 14.69s/it]\n",
      "100%|██████████| 13460/13460 [01:56<00:00, 115.55it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.8653938523497438\n",
      "==========================================\n",
      "Combination 31\n",
      "==========================================\n",
      "Factors = 25\n",
      "Regularization = 1\n",
      "Alpha = 40\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [02:30<00:00, 32.28s/it]\n",
      "100%|██████████| 13460/13460 [02:12<00:00, 101.33it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.854681478686493\n",
      "==========================================\n",
      "Combination 32\n",
      "==========================================\n",
      "Factors = 25\n",
      "Regularization = 0.01\n",
      "Alpha = 40\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [02:26<00:00, 32.14s/it]\n",
      "100%|██████████| 13460/13460 [02:14<00:00, 100.01it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.8543539364788071\n",
      "==========================================\n",
      "Combination 33\n",
      "==========================================\n",
      "Factors = 25\n",
      "Regularization = 0.0001\n",
      "Alpha = 40\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [02:28<00:00, 33.06s/it]\n",
      "100%|██████████| 13460/13460 [02:12<00:00, 101.65it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.8555122540772739\n",
      "==========================================\n",
      "Combination 34\n",
      "==========================================\n",
      "Factors = 50\n",
      "Regularization = 1\n",
      "Alpha = 40\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [03:37<00:00, 46.92s/it]\n",
      "100%|██████████| 13460/13460 [02:15<00:00, 99.37it/s] \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.8436847699365719\n",
      "==========================================\n",
      "Combination 35\n",
      "==========================================\n",
      "Factors = 50\n",
      "Regularization = 0.01\n",
      "Alpha = 40\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [03:30<00:00, 47.32s/it]\n",
      "100%|██████████| 13460/13460 [02:13<00:00, 100.98it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.844001177659708\n",
      "==========================================\n",
      "Combination 36\n",
      "==========================================\n",
      "Factors = 50\n",
      "Regularization = 0.0001\n",
      "Alpha = 40\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [03:27<00:00, 44.67s/it]\n",
      "100%|██████████| 13460/13460 [02:18<00:00, 97.33it/s] "
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.8432389334531324\n",
      "==========================================\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>factors</th>\n",
       "      <th>regularization</th>\n",
       "      <th>alpha</th>\n",
       "      <th>auc</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>25.0</td>\n",
       "      <td>0.0001</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.870082</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>25.0</td>\n",
       "      <td>1.0000</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.868217</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>10.0</td>\n",
       "      <td>1.0000</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.867958</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>10.0</td>\n",
       "      <td>0.0001</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.867649</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>10.0</td>\n",
       "      <td>0.0100</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.867570</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>10.0</td>\n",
       "      <td>1.0000</td>\n",
       "      <td>15.0</td>\n",
       "      <td>0.867010</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>25.0</td>\n",
       "      <td>0.0100</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.866539</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>10.0</td>\n",
       "      <td>0.0100</td>\n",
       "      <td>15.0</td>\n",
       "      <td>0.866039</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>29</th>\n",
       "      <td>10.0</td>\n",
       "      <td>0.0001</td>\n",
       "      <td>40.0</td>\n",
       "      <td>0.865394</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>10.0</td>\n",
       "      <td>0.0001</td>\n",
       "      <td>15.0</td>\n",
       "      <td>0.865114</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>10.0</td>\n",
       "      <td>0.0100</td>\n",
       "      <td>30.0</td>\n",
       "      <td>0.865035</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>10.0</td>\n",
       "      <td>1.0000</td>\n",
       "      <td>40.0</td>\n",
       "      <td>0.864732</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>10.0</td>\n",
       "      <td>0.0001</td>\n",
       "      <td>30.0</td>\n",
       "      <td>0.864139</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>28</th>\n",
       "      <td>10.0</td>\n",
       "      <td>0.0100</td>\n",
       "      <td>40.0</td>\n",
       "      <td>0.863873</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>10.0</td>\n",
       "      <td>1.0000</td>\n",
       "      <td>30.0</td>\n",
       "      <td>0.863563</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>25.0</td>\n",
       "      <td>0.0100</td>\n",
       "      <td>15.0</td>\n",
       "      <td>0.861417</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>50.0</td>\n",
       "      <td>0.0001</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.861343</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>50.0</td>\n",
       "      <td>0.0100</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.860394</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>50.0</td>\n",
       "      <td>1.0000</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.859469</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>25.0</td>\n",
       "      <td>1.0000</td>\n",
       "      <td>15.0</td>\n",
       "      <td>0.859159</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>25.0</td>\n",
       "      <td>0.0001</td>\n",
       "      <td>15.0</td>\n",
       "      <td>0.858931</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>25.0</td>\n",
       "      <td>0.0100</td>\n",
       "      <td>30.0</td>\n",
       "      <td>0.856522</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>25.0</td>\n",
       "      <td>0.0001</td>\n",
       "      <td>30.0</td>\n",
       "      <td>0.856226</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21</th>\n",
       "      <td>25.0</td>\n",
       "      <td>1.0000</td>\n",
       "      <td>30.0</td>\n",
       "      <td>0.855834</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>32</th>\n",
       "      <td>25.0</td>\n",
       "      <td>0.0001</td>\n",
       "      <td>40.0</td>\n",
       "      <td>0.855512</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30</th>\n",
       "      <td>25.0</td>\n",
       "      <td>1.0000</td>\n",
       "      <td>40.0</td>\n",
       "      <td>0.854681</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>31</th>\n",
       "      <td>25.0</td>\n",
       "      <td>0.0100</td>\n",
       "      <td>40.0</td>\n",
       "      <td>0.854354</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>50.0</td>\n",
       "      <td>1.0000</td>\n",
       "      <td>15.0</td>\n",
       "      <td>0.849484</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>50.0</td>\n",
       "      <td>0.0100</td>\n",
       "      <td>15.0</td>\n",
       "      <td>0.848292</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26</th>\n",
       "      <td>50.0</td>\n",
       "      <td>0.0001</td>\n",
       "      <td>30.0</td>\n",
       "      <td>0.846366</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>50.0</td>\n",
       "      <td>0.0001</td>\n",
       "      <td>15.0</td>\n",
       "      <td>0.845909</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24</th>\n",
       "      <td>50.0</td>\n",
       "      <td>1.0000</td>\n",
       "      <td>30.0</td>\n",
       "      <td>0.845448</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>50.0</td>\n",
       "      <td>0.0100</td>\n",
       "      <td>30.0</td>\n",
       "      <td>0.844548</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>34</th>\n",
       "      <td>50.0</td>\n",
       "      <td>0.0100</td>\n",
       "      <td>40.0</td>\n",
       "      <td>0.844001</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>33</th>\n",
       "      <td>50.0</td>\n",
       "      <td>1.0000</td>\n",
       "      <td>40.0</td>\n",
       "      <td>0.843685</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35</th>\n",
       "      <td>50.0</td>\n",
       "      <td>0.0001</td>\n",
       "      <td>40.0</td>\n",
       "      <td>0.843239</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    factors  regularization  alpha       auc\n",
       "5      25.0          0.0001    1.0  0.870082\n",
       "3      25.0          1.0000    1.0  0.868217\n",
       "0      10.0          1.0000    1.0  0.867958\n",
       "2      10.0          0.0001    1.0  0.867649\n",
       "1      10.0          0.0100    1.0  0.867570\n",
       "9      10.0          1.0000   15.0  0.867010\n",
       "4      25.0          0.0100    1.0  0.866539\n",
       "10     10.0          0.0100   15.0  0.866039\n",
       "29     10.0          0.0001   40.0  0.865394\n",
       "11     10.0          0.0001   15.0  0.865114\n",
       "19     10.0          0.0100   30.0  0.865035\n",
       "27     10.0          1.0000   40.0  0.864732\n",
       "20     10.0          0.0001   30.0  0.864139\n",
       "28     10.0          0.0100   40.0  0.863873\n",
       "18     10.0          1.0000   30.0  0.863563\n",
       "13     25.0          0.0100   15.0  0.861417\n",
       "8      50.0          0.0001    1.0  0.861343\n",
       "7      50.0          0.0100    1.0  0.860394\n",
       "6      50.0          1.0000    1.0  0.859469\n",
       "12     25.0          1.0000   15.0  0.859159\n",
       "14     25.0          0.0001   15.0  0.858931\n",
       "22     25.0          0.0100   30.0  0.856522\n",
       "23     25.0          0.0001   30.0  0.856226\n",
       "21     25.0          1.0000   30.0  0.855834\n",
       "32     25.0          0.0001   40.0  0.855512\n",
       "30     25.0          1.0000   40.0  0.854681\n",
       "31     25.0          0.0100   40.0  0.854354\n",
       "15     50.0          1.0000   15.0  0.849484\n",
       "16     50.0          0.0100   15.0  0.848292\n",
       "26     50.0          0.0001   30.0  0.846366\n",
       "17     50.0          0.0001   15.0  0.845909\n",
       "24     50.0          1.0000   30.0  0.845448\n",
       "25     50.0          0.0100   30.0  0.844548\n",
       "34     50.0          0.0100   40.0  0.844001\n",
       "33     50.0          1.0000   40.0  0.843685\n",
       "35     50.0          0.0001   40.0  0.843239"
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.model_selection import ParameterGrid\n",
    "\n",
    "# Define the hyperparameters\n",
    "hyperparameters = {\n",
    "    'factors': [10, 25, 50], \n",
    "    'regularization' : [1, 0.01,0.0001], \n",
    "    'alpha' : [1, 15, 30, 40]}\n",
    "\n",
    "# Compute all combinations\n",
    "parameter_grid = list(ParameterGrid(hyperparameters))\n",
    "\n",
    "# Just a table to save the results\n",
    "resultsDF = pd.DataFrame(columns=['factors', 'regularization', 'alpha', 'auc'])\n",
    "\n",
    "iter = 1\n",
    "\n",
    "for parameter_combination in parameter_grid:    \n",
    "    print(\"Combination\", iter)\n",
    "    print(\"==========================================\")\n",
    "    iter += 1\n",
    "    print(\"Factors =\",parameter_combination['factors'])\n",
    "    print(\"Regularization =\",parameter_combination['regularization'])\n",
    "    print(\"Alpha =\",parameter_combination['alpha'])\n",
    "\n",
    "    # Create the model\n",
    "    model = implicit.als.AlternatingLeastSquares(factors=parameter_combination['factors'], regularization=parameter_combination['regularization'], iterations=5)\n",
    "\n",
    "    # Apply alpha to the item_user matrix\n",
    "    train_data_conf = (train_sparse_item_user * parameter_combination['alpha'] ).astype('double')\n",
    "\n",
    "    # Fit the model\n",
    "    model.fit(train_data_conf)\n",
    "    \n",
    "    # Compute AUC distribution\n",
    "    # 10% ratio was used because the calculation of auc for 100% of test users takes more than 20 mins \n",
    "    auc = compute_auc(test_users_ratio = 0.1)\n",
    "    \n",
    "    parameter_combination['auc'] = auc.mean()\n",
    "    \n",
    "    resultsDF = resultsDF.append(parameter_combination, ignore_index=True)\n",
    "    \n",
    "    print(\"Average AUC = \", parameter_combination['auc'])\n",
    "    print(\"==========================================\")\n",
    "    \n",
    "resultsDF.sort_values('auc', ascending = False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>factors</th>\n",
       "      <th>regularization</th>\n",
       "      <th>alpha</th>\n",
       "      <th>auc</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>25.0</td>\n",
       "      <td>0.0001</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.870082</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>25.0</td>\n",
       "      <td>1.0000</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.868217</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>10.0</td>\n",
       "      <td>1.0000</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.867958</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>10.0</td>\n",
       "      <td>0.0001</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.867649</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>10.0</td>\n",
       "      <td>0.0100</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.867570</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>10.0</td>\n",
       "      <td>1.0000</td>\n",
       "      <td>15.0</td>\n",
       "      <td>0.867010</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>25.0</td>\n",
       "      <td>0.0100</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.866539</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>10.0</td>\n",
       "      <td>0.0100</td>\n",
       "      <td>15.0</td>\n",
       "      <td>0.866039</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>29</th>\n",
       "      <td>10.0</td>\n",
       "      <td>0.0001</td>\n",
       "      <td>40.0</td>\n",
       "      <td>0.865394</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>10.0</td>\n",
       "      <td>0.0001</td>\n",
       "      <td>15.0</td>\n",
       "      <td>0.865114</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>10.0</td>\n",
       "      <td>0.0100</td>\n",
       "      <td>30.0</td>\n",
       "      <td>0.865035</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>10.0</td>\n",
       "      <td>1.0000</td>\n",
       "      <td>40.0</td>\n",
       "      <td>0.864732</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>10.0</td>\n",
       "      <td>0.0001</td>\n",
       "      <td>30.0</td>\n",
       "      <td>0.864139</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>28</th>\n",
       "      <td>10.0</td>\n",
       "      <td>0.0100</td>\n",
       "      <td>40.0</td>\n",
       "      <td>0.863873</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>10.0</td>\n",
       "      <td>1.0000</td>\n",
       "      <td>30.0</td>\n",
       "      <td>0.863563</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>25.0</td>\n",
       "      <td>0.0100</td>\n",
       "      <td>15.0</td>\n",
       "      <td>0.861417</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>50.0</td>\n",
       "      <td>0.0001</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.861343</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>50.0</td>\n",
       "      <td>0.0100</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.860394</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>50.0</td>\n",
       "      <td>1.0000</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.859469</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>25.0</td>\n",
       "      <td>1.0000</td>\n",
       "      <td>15.0</td>\n",
       "      <td>0.859159</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>25.0</td>\n",
       "      <td>0.0001</td>\n",
       "      <td>15.0</td>\n",
       "      <td>0.858931</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>25.0</td>\n",
       "      <td>0.0100</td>\n",
       "      <td>30.0</td>\n",
       "      <td>0.856522</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>25.0</td>\n",
       "      <td>0.0001</td>\n",
       "      <td>30.0</td>\n",
       "      <td>0.856226</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21</th>\n",
       "      <td>25.0</td>\n",
       "      <td>1.0000</td>\n",
       "      <td>30.0</td>\n",
       "      <td>0.855834</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>32</th>\n",
       "      <td>25.0</td>\n",
       "      <td>0.0001</td>\n",
       "      <td>40.0</td>\n",
       "      <td>0.855512</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30</th>\n",
       "      <td>25.0</td>\n",
       "      <td>1.0000</td>\n",
       "      <td>40.0</td>\n",
       "      <td>0.854681</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>31</th>\n",
       "      <td>25.0</td>\n",
       "      <td>0.0100</td>\n",
       "      <td>40.0</td>\n",
       "      <td>0.854354</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>50.0</td>\n",
       "      <td>1.0000</td>\n",
       "      <td>15.0</td>\n",
       "      <td>0.849484</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>50.0</td>\n",
       "      <td>0.0100</td>\n",
       "      <td>15.0</td>\n",
       "      <td>0.848292</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26</th>\n",
       "      <td>50.0</td>\n",
       "      <td>0.0001</td>\n",
       "      <td>30.0</td>\n",
       "      <td>0.846366</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>50.0</td>\n",
       "      <td>0.0001</td>\n",
       "      <td>15.0</td>\n",
       "      <td>0.845909</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24</th>\n",
       "      <td>50.0</td>\n",
       "      <td>1.0000</td>\n",
       "      <td>30.0</td>\n",
       "      <td>0.845448</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>50.0</td>\n",
       "      <td>0.0100</td>\n",
       "      <td>30.0</td>\n",
       "      <td>0.844548</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>34</th>\n",
       "      <td>50.0</td>\n",
       "      <td>0.0100</td>\n",
       "      <td>40.0</td>\n",
       "      <td>0.844001</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>33</th>\n",
       "      <td>50.0</td>\n",
       "      <td>1.0000</td>\n",
       "      <td>40.0</td>\n",
       "      <td>0.843685</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35</th>\n",
       "      <td>50.0</td>\n",
       "      <td>0.0001</td>\n",
       "      <td>40.0</td>\n",
       "      <td>0.843239</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    factors  regularization  alpha       auc\n",
       "5      25.0          0.0001    1.0  0.870082\n",
       "3      25.0          1.0000    1.0  0.868217\n",
       "0      10.0          1.0000    1.0  0.867958\n",
       "2      10.0          0.0001    1.0  0.867649\n",
       "1      10.0          0.0100    1.0  0.867570\n",
       "9      10.0          1.0000   15.0  0.867010\n",
       "4      25.0          0.0100    1.0  0.866539\n",
       "10     10.0          0.0100   15.0  0.866039\n",
       "29     10.0          0.0001   40.0  0.865394\n",
       "11     10.0          0.0001   15.0  0.865114\n",
       "19     10.0          0.0100   30.0  0.865035\n",
       "27     10.0          1.0000   40.0  0.864732\n",
       "20     10.0          0.0001   30.0  0.864139\n",
       "28     10.0          0.0100   40.0  0.863873\n",
       "18     10.0          1.0000   30.0  0.863563\n",
       "13     25.0          0.0100   15.0  0.861417\n",
       "8      50.0          0.0001    1.0  0.861343\n",
       "7      50.0          0.0100    1.0  0.860394\n",
       "6      50.0          1.0000    1.0  0.859469\n",
       "12     25.0          1.0000   15.0  0.859159\n",
       "14     25.0          0.0001   15.0  0.858931\n",
       "22     25.0          0.0100   30.0  0.856522\n",
       "23     25.0          0.0001   30.0  0.856226\n",
       "21     25.0          1.0000   30.0  0.855834\n",
       "32     25.0          0.0001   40.0  0.855512\n",
       "30     25.0          1.0000   40.0  0.854681\n",
       "31     25.0          0.0100   40.0  0.854354\n",
       "15     50.0          1.0000   15.0  0.849484\n",
       "16     50.0          0.0100   15.0  0.848292\n",
       "26     50.0          0.0001   30.0  0.846366\n",
       "17     50.0          0.0001   15.0  0.845909\n",
       "24     50.0          1.0000   30.0  0.845448\n",
       "25     50.0          0.0100   30.0  0.844548\n",
       "34     50.0          0.0100   40.0  0.844001\n",
       "33     50.0          1.0000   40.0  0.843685\n",
       "35     50.0          0.0001   40.0  0.843239"
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "resultsDF = resultsDF.sort_values('auc', ascending = False)\n",
    "resultsDF"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### Question 10.2\n",
    "\n",
    "Sometimes an exhaustive search in the hyperparameter space is a waste of time and resources.\n",
    "There are a lot of research questions open that deal with automating the end-to-end process of applying machine learning to real-world problems (e.g. **AutoML**).\n",
    "\n",
    "Random sampling in the hyperparameter space is a possible -- very simple but yet quite powerfull -- technique.\n",
    "\n",
    "Implement a simple parameter sampler and try a bunch of different configurations.\n",
    "\n",
    "***Hint***: have a look to `ParameterSampler()` in `sklearn.model_selection` "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {
    "scrolled": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Combination 37\n",
      "==========================================\n",
      "Factors = 47\n",
      "Regularization = 0.42119743435482593\n",
      "Alpha = 3.3256130350593374\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [03:09<00:00, 41.32s/it]\n",
      "100%|██████████| 13460/13460 [02:17<00:00, 97.78it/s] \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.8562689132890848\n",
      "==========================================\n",
      "Combination 38\n",
      "==========================================\n",
      "Factors = 31\n",
      "Regularization = 0.013217574417221722\n",
      "Alpha = 38.6623228652102\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [02:30<00:00, 32.63s/it]\n",
      "100%|██████████| 13460/13460 [02:16<00:00, 98.51it/s] \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.8505830263628311\n",
      "==========================================\n",
      "Combination 39\n",
      "==========================================\n",
      "Factors = 24\n",
      "Regularization = 0.3819369943970108\n",
      "Alpha = 23.055279869638657\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [02:07<00:00, 28.28s/it]\n",
      "100%|██████████| 13460/13460 [02:15<00:00, 99.08it/s] \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.8594750974140464\n",
      "==========================================\n",
      "Combination 40\n",
      "==========================================\n",
      "Factors = 32\n",
      "Regularization = 0.7776179787050381\n",
      "Alpha = 10.383272838714394\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [02:06<00:00, 27.66s/it]\n",
      "100%|██████████| 13460/13460 [02:09<00:00, 104.04it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.8584786046082057\n",
      "==========================================\n",
      "Combination 41\n",
      "==========================================\n",
      "Factors = 30\n",
      "Regularization = 0.1046259474719279\n",
      "Alpha = 30.460252740693498\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [02:32<00:00, 32.95s/it]\n",
      "100%|██████████| 13460/13460 [02:16<00:00, 98.28it/s] \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.8549759520344696\n",
      "==========================================\n",
      "Combination 42\n",
      "==========================================\n",
      "Factors = 48\n",
      "Regularization = 0.8186496505993194\n",
      "Alpha = 33.26413210025606\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [03:05<00:00, 40.44s/it]\n",
      "100%|██████████| 13460/13460 [02:14<00:00, 100.44it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.8474135385373943\n",
      "==========================================\n",
      "Combination 43\n",
      "==========================================\n",
      "Factors = 40\n",
      "Regularization = 0.053052241357943744\n",
      "Alpha = 40.12990615265874\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [02:37<00:00, 35.59s/it]\n",
      "100%|██████████| 13460/13460 [02:14<00:00, 100.08it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.8516436346468906\n",
      "==========================================\n",
      "Combination 44\n",
      "==========================================\n",
      "Factors = 31\n",
      "Regularization = 0.5887050687472363\n",
      "Alpha = 17.116074207072494\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [02:39<00:00, 34.73s/it]\n",
      "100%|██████████| 13460/13460 [02:13<00:00, 100.91it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.8556486806479168\n",
      "==========================================\n",
      "Combination 45\n",
      "==========================================\n",
      "Factors = 22\n",
      "Regularization = 0.6358296629077513\n",
      "Alpha = 23.529467384309854\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [02:08<00:00, 28.06s/it]\n",
      "100%|██████████| 13460/13460 [02:12<00:00, 101.91it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.8581941422683669\n",
      "==========================================\n",
      "Combination 46\n",
      "==========================================\n",
      "Factors = 19\n",
      "Regularization = 0.5749603526375375\n",
      "Alpha = 5.197703869519018\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [01:52<00:00, 25.08s/it]\n",
      "100%|██████████| 13460/13460 [02:18<00:00, 97.50it/s] \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.8623566839003826\n",
      "==========================================\n",
      "Combination 47\n",
      "==========================================\n",
      "Factors = 23\n",
      "Regularization = 0.5628305296353057\n",
      "Alpha = 8.646115197662372\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [02:08<00:00, 28.74s/it]\n",
      "100%|██████████| 13460/13460 [02:20<00:00, 95.86it/s] \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.8610572054358611\n",
      "==========================================\n",
      "Combination 48\n",
      "==========================================\n",
      "Factors = 28\n",
      "Regularization = 0.5732089438078\n",
      "Alpha = 35.403291462994\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5.0/5 [02:29<00:00, 33.00s/it]\n",
      "100%|██████████| 13460/13460 [02:12<00:00, 101.22it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average AUC =  0.8537814043250005\n",
      "==========================================\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>factors</th>\n",
       "      <th>regularization</th>\n",
       "      <th>alpha</th>\n",
       "      <th>auc</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>25.0</td>\n",
       "      <td>0.000100</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.870082</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>25.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.868217</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>10.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.867958</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>10.0</td>\n",
       "      <td>0.000100</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.867649</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>10.0</td>\n",
       "      <td>0.010000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.867570</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>10.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>15.000000</td>\n",
       "      <td>0.867010</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>25.0</td>\n",
       "      <td>0.010000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.866539</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>10.0</td>\n",
       "      <td>0.010000</td>\n",
       "      <td>15.000000</td>\n",
       "      <td>0.866039</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>10.0</td>\n",
       "      <td>0.000100</td>\n",
       "      <td>40.000000</td>\n",
       "      <td>0.865394</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>10.0</td>\n",
       "      <td>0.000100</td>\n",
       "      <td>15.000000</td>\n",
       "      <td>0.865114</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>10.0</td>\n",
       "      <td>0.010000</td>\n",
       "      <td>30.000000</td>\n",
       "      <td>0.865035</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>10.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>40.000000</td>\n",
       "      <td>0.864732</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>10.0</td>\n",
       "      <td>0.000100</td>\n",
       "      <td>30.000000</td>\n",
       "      <td>0.864139</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>10.0</td>\n",
       "      <td>0.010000</td>\n",
       "      <td>40.000000</td>\n",
       "      <td>0.863873</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>10.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>30.000000</td>\n",
       "      <td>0.863563</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>45</th>\n",
       "      <td>19.0</td>\n",
       "      <td>0.574960</td>\n",
       "      <td>5.197704</td>\n",
       "      <td>0.862357</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>25.0</td>\n",
       "      <td>0.010000</td>\n",
       "      <td>15.000000</td>\n",
       "      <td>0.861417</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>50.0</td>\n",
       "      <td>0.000100</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.861343</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>46</th>\n",
       "      <td>23.0</td>\n",
       "      <td>0.562831</td>\n",
       "      <td>8.646115</td>\n",
       "      <td>0.861057</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>50.0</td>\n",
       "      <td>0.010000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.860394</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>38</th>\n",
       "      <td>24.0</td>\n",
       "      <td>0.381937</td>\n",
       "      <td>23.055280</td>\n",
       "      <td>0.859475</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>50.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.859469</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>25.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>15.000000</td>\n",
       "      <td>0.859159</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>25.0</td>\n",
       "      <td>0.000100</td>\n",
       "      <td>15.000000</td>\n",
       "      <td>0.858931</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>39</th>\n",
       "      <td>32.0</td>\n",
       "      <td>0.777618</td>\n",
       "      <td>10.383273</td>\n",
       "      <td>0.858479</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>44</th>\n",
       "      <td>22.0</td>\n",
       "      <td>0.635830</td>\n",
       "      <td>23.529467</td>\n",
       "      <td>0.858194</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21</th>\n",
       "      <td>25.0</td>\n",
       "      <td>0.010000</td>\n",
       "      <td>30.000000</td>\n",
       "      <td>0.856522</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>36</th>\n",
       "      <td>47.0</td>\n",
       "      <td>0.421197</td>\n",
       "      <td>3.325613</td>\n",
       "      <td>0.856269</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>25.0</td>\n",
       "      <td>0.000100</td>\n",
       "      <td>30.000000</td>\n",
       "      <td>0.856226</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>25.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>30.000000</td>\n",
       "      <td>0.855834</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>43</th>\n",
       "      <td>31.0</td>\n",
       "      <td>0.588705</td>\n",
       "      <td>17.116074</td>\n",
       "      <td>0.855649</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24</th>\n",
       "      <td>25.0</td>\n",
       "      <td>0.000100</td>\n",
       "      <td>40.000000</td>\n",
       "      <td>0.855512</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>40</th>\n",
       "      <td>30.0</td>\n",
       "      <td>0.104626</td>\n",
       "      <td>30.460253</td>\n",
       "      <td>0.854976</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>25.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>40.000000</td>\n",
       "      <td>0.854681</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26</th>\n",
       "      <td>25.0</td>\n",
       "      <td>0.010000</td>\n",
       "      <td>40.000000</td>\n",
       "      <td>0.854354</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>47</th>\n",
       "      <td>28.0</td>\n",
       "      <td>0.573209</td>\n",
       "      <td>35.403291</td>\n",
       "      <td>0.853781</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>42</th>\n",
       "      <td>40.0</td>\n",
       "      <td>0.053052</td>\n",
       "      <td>40.129906</td>\n",
       "      <td>0.851644</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>37</th>\n",
       "      <td>31.0</td>\n",
       "      <td>0.013218</td>\n",
       "      <td>38.662323</td>\n",
       "      <td>0.850583</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>50.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>15.000000</td>\n",
       "      <td>0.849484</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>28</th>\n",
       "      <td>50.0</td>\n",
       "      <td>0.010000</td>\n",
       "      <td>15.000000</td>\n",
       "      <td>0.848292</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>41</th>\n",
       "      <td>48.0</td>\n",
       "      <td>0.818650</td>\n",
       "      <td>33.264132</td>\n",
       "      <td>0.847414</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>29</th>\n",
       "      <td>50.0</td>\n",
       "      <td>0.000100</td>\n",
       "      <td>30.000000</td>\n",
       "      <td>0.846366</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30</th>\n",
       "      <td>50.0</td>\n",
       "      <td>0.000100</td>\n",
       "      <td>15.000000</td>\n",
       "      <td>0.845909</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>31</th>\n",
       "      <td>50.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>30.000000</td>\n",
       "      <td>0.845448</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>32</th>\n",
       "      <td>50.0</td>\n",
       "      <td>0.010000</td>\n",
       "      <td>30.000000</td>\n",
       "      <td>0.844548</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>33</th>\n",
       "      <td>50.0</td>\n",
       "      <td>0.010000</td>\n",
       "      <td>40.000000</td>\n",
       "      <td>0.844001</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>34</th>\n",
       "      <td>50.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>40.000000</td>\n",
       "      <td>0.843685</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35</th>\n",
       "      <td>50.0</td>\n",
       "      <td>0.000100</td>\n",
       "      <td>40.000000</td>\n",
       "      <td>0.843239</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    factors  regularization      alpha       auc\n",
       "0      25.0        0.000100   1.000000  0.870082\n",
       "1      25.0        1.000000   1.000000  0.868217\n",
       "2      10.0        1.000000   1.000000  0.867958\n",
       "3      10.0        0.000100   1.000000  0.867649\n",
       "4      10.0        0.010000   1.000000  0.867570\n",
       "5      10.0        1.000000  15.000000  0.867010\n",
       "6      25.0        0.010000   1.000000  0.866539\n",
       "7      10.0        0.010000  15.000000  0.866039\n",
       "8      10.0        0.000100  40.000000  0.865394\n",
       "9      10.0        0.000100  15.000000  0.865114\n",
       "10     10.0        0.010000  30.000000  0.865035\n",
       "11     10.0        1.000000  40.000000  0.864732\n",
       "12     10.0        0.000100  30.000000  0.864139\n",
       "13     10.0        0.010000  40.000000  0.863873\n",
       "14     10.0        1.000000  30.000000  0.863563\n",
       "45     19.0        0.574960   5.197704  0.862357\n",
       "15     25.0        0.010000  15.000000  0.861417\n",
       "16     50.0        0.000100   1.000000  0.861343\n",
       "46     23.0        0.562831   8.646115  0.861057\n",
       "17     50.0        0.010000   1.000000  0.860394\n",
       "38     24.0        0.381937  23.055280  0.859475\n",
       "18     50.0        1.000000   1.000000  0.859469\n",
       "19     25.0        1.000000  15.000000  0.859159\n",
       "20     25.0        0.000100  15.000000  0.858931\n",
       "39     32.0        0.777618  10.383273  0.858479\n",
       "44     22.0        0.635830  23.529467  0.858194\n",
       "21     25.0        0.010000  30.000000  0.856522\n",
       "36     47.0        0.421197   3.325613  0.856269\n",
       "22     25.0        0.000100  30.000000  0.856226\n",
       "23     25.0        1.000000  30.000000  0.855834\n",
       "43     31.0        0.588705  17.116074  0.855649\n",
       "24     25.0        0.000100  40.000000  0.855512\n",
       "40     30.0        0.104626  30.460253  0.854976\n",
       "25     25.0        1.000000  40.000000  0.854681\n",
       "26     25.0        0.010000  40.000000  0.854354\n",
       "47     28.0        0.573209  35.403291  0.853781\n",
       "42     40.0        0.053052  40.129906  0.851644\n",
       "37     31.0        0.013218  38.662323  0.850583\n",
       "27     50.0        1.000000  15.000000  0.849484\n",
       "28     50.0        0.010000  15.000000  0.848292\n",
       "41     48.0        0.818650  33.264132  0.847414\n",
       "29     50.0        0.000100  30.000000  0.846366\n",
       "30     50.0        0.000100  15.000000  0.845909\n",
       "31     50.0        1.000000  30.000000  0.845448\n",
       "32     50.0        0.010000  30.000000  0.844548\n",
       "33     50.0        0.010000  40.000000  0.844001\n",
       "34     50.0        1.000000  40.000000  0.843685\n",
       "35     50.0        0.000100  40.000000  0.843239"
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import scipy.stats \n",
    "from sklearn.model_selection import ParameterSampler\n",
    "\n",
    "# Hyperparameters are now definded by probability densities\n",
    "hyperparameters = {\n",
    "    'factors': scipy.stats.randint(10, 50), \n",
    "    'regularization' : scipy.stats.uniform(0.0001, 1.), \n",
    "    'alpha' : scipy.stats.uniform(1., 40.)}\n",
    "\n",
    "parameter_grid = list(ParameterSampler(hyperparameters, 12))\n",
    "\n",
    "# You can either append the results to the previous table or create a new one\n",
    "# resultsDF_sampling = pd.DataFrame(columns=['factors', 'regularization', 'alpha', 'auc'])\n",
    "\n",
    "for parameter_combination in parameter_grid:  \n",
    "    \n",
    "    print(\"Combination\", iter)\n",
    "    print(\"==========================================\")\n",
    "    iter += 1\n",
    "    print(\"Factors =\",parameter_combination['factors'])\n",
    "    print(\"Regularization =\",parameter_combination['regularization'])\n",
    "    print(\"Alpha =\",parameter_combination['alpha'])\n",
    "    \n",
    "    # Create the model\n",
    "    model = implicit.als.AlternatingLeastSquares(factors=parameter_combination['factors'], regularization=parameter_combination['regularization'], iterations=5)\n",
    "    \n",
    "    # Apply alpha to the item_user matrix\n",
    "    train_data_conf = (train_sparse_item_user * parameter_combination['alpha'] ).astype('double')\n",
    "\n",
    "    # Fit the model\n",
    "    model.fit(train_data_conf)\n",
    "    \n",
    "    # Compute AUC distribution\n",
    "    auc = compute_auc(test_users_ratio = 0.1)\n",
    "    \n",
    "    parameter_combination['auc'] = auc.mean()\n",
    "    \n",
    "    resultsDF = resultsDF.append(parameter_combination, ignore_index=True)\n",
    "    \n",
    "    print(\"Average AUC = \", parameter_combination['auc'])\n",
    "    print(\"==========================================\")\n",
    " \n",
    "    \n",
    "resultsDF.sort_values('auc', ascending = False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>factors</th>\n",
       "      <th>regularization</th>\n",
       "      <th>alpha</th>\n",
       "      <th>auc</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>25.0</td>\n",
       "      <td>0.000100</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.870082</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>25.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.868217</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>10.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.867958</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>10.0</td>\n",
       "      <td>0.000100</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.867649</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>10.0</td>\n",
       "      <td>0.010000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.867570</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>10.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>15.000000</td>\n",
       "      <td>0.867010</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>25.0</td>\n",
       "      <td>0.010000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.866539</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>10.0</td>\n",
       "      <td>0.010000</td>\n",
       "      <td>15.000000</td>\n",
       "      <td>0.866039</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>10.0</td>\n",
       "      <td>0.000100</td>\n",
       "      <td>40.000000</td>\n",
       "      <td>0.865394</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>10.0</td>\n",
       "      <td>0.000100</td>\n",
       "      <td>15.000000</td>\n",
       "      <td>0.865114</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>10.0</td>\n",
       "      <td>0.010000</td>\n",
       "      <td>30.000000</td>\n",
       "      <td>0.865035</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>10.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>40.000000</td>\n",
       "      <td>0.864732</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>10.0</td>\n",
       "      <td>0.000100</td>\n",
       "      <td>30.000000</td>\n",
       "      <td>0.864139</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>10.0</td>\n",
       "      <td>0.010000</td>\n",
       "      <td>40.000000</td>\n",
       "      <td>0.863873</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>10.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>30.000000</td>\n",
       "      <td>0.863563</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>45</th>\n",
       "      <td>19.0</td>\n",
       "      <td>0.574960</td>\n",
       "      <td>5.197704</td>\n",
       "      <td>0.862357</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>25.0</td>\n",
       "      <td>0.010000</td>\n",
       "      <td>15.000000</td>\n",
       "      <td>0.861417</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>50.0</td>\n",
       "      <td>0.000100</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.861343</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>46</th>\n",
       "      <td>23.0</td>\n",
       "      <td>0.562831</td>\n",
       "      <td>8.646115</td>\n",
       "      <td>0.861057</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>50.0</td>\n",
       "      <td>0.010000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.860394</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>38</th>\n",
       "      <td>24.0</td>\n",
       "      <td>0.381937</td>\n",
       "      <td>23.055280</td>\n",
       "      <td>0.859475</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>50.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.859469</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>25.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>15.000000</td>\n",
       "      <td>0.859159</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>25.0</td>\n",
       "      <td>0.000100</td>\n",
       "      <td>15.000000</td>\n",
       "      <td>0.858931</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>39</th>\n",
       "      <td>32.0</td>\n",
       "      <td>0.777618</td>\n",
       "      <td>10.383273</td>\n",
       "      <td>0.858479</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>44</th>\n",
       "      <td>22.0</td>\n",
       "      <td>0.635830</td>\n",
       "      <td>23.529467</td>\n",
       "      <td>0.858194</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21</th>\n",
       "      <td>25.0</td>\n",
       "      <td>0.010000</td>\n",
       "      <td>30.000000</td>\n",
       "      <td>0.856522</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>36</th>\n",
       "      <td>47.0</td>\n",
       "      <td>0.421197</td>\n",
       "      <td>3.325613</td>\n",
       "      <td>0.856269</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>25.0</td>\n",
       "      <td>0.000100</td>\n",
       "      <td>30.000000</td>\n",
       "      <td>0.856226</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>25.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>30.000000</td>\n",
       "      <td>0.855834</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>43</th>\n",
       "      <td>31.0</td>\n",
       "      <td>0.588705</td>\n",
       "      <td>17.116074</td>\n",
       "      <td>0.855649</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24</th>\n",
       "      <td>25.0</td>\n",
       "      <td>0.000100</td>\n",
       "      <td>40.000000</td>\n",
       "      <td>0.855512</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>40</th>\n",
       "      <td>30.0</td>\n",
       "      <td>0.104626</td>\n",
       "      <td>30.460253</td>\n",
       "      <td>0.854976</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>25.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>40.000000</td>\n",
       "      <td>0.854681</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26</th>\n",
       "      <td>25.0</td>\n",
       "      <td>0.010000</td>\n",
       "      <td>40.000000</td>\n",
       "      <td>0.854354</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>47</th>\n",
       "      <td>28.0</td>\n",
       "      <td>0.573209</td>\n",
       "      <td>35.403291</td>\n",
       "      <td>0.853781</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>42</th>\n",
       "      <td>40.0</td>\n",
       "      <td>0.053052</td>\n",
       "      <td>40.129906</td>\n",
       "      <td>0.851644</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>37</th>\n",
       "      <td>31.0</td>\n",
       "      <td>0.013218</td>\n",
       "      <td>38.662323</td>\n",
       "      <td>0.850583</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>50.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>15.000000</td>\n",
       "      <td>0.849484</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>28</th>\n",
       "      <td>50.0</td>\n",
       "      <td>0.010000</td>\n",
       "      <td>15.000000</td>\n",
       "      <td>0.848292</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>41</th>\n",
       "      <td>48.0</td>\n",
       "      <td>0.818650</td>\n",
       "      <td>33.264132</td>\n",
       "      <td>0.847414</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>29</th>\n",
       "      <td>50.0</td>\n",
       "      <td>0.000100</td>\n",
       "      <td>30.000000</td>\n",
       "      <td>0.846366</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30</th>\n",
       "      <td>50.0</td>\n",
       "      <td>0.000100</td>\n",
       "      <td>15.000000</td>\n",
       "      <td>0.845909</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>31</th>\n",
       "      <td>50.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>30.000000</td>\n",
       "      <td>0.845448</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>32</th>\n",
       "      <td>50.0</td>\n",
       "      <td>0.010000</td>\n",
       "      <td>30.000000</td>\n",
       "      <td>0.844548</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>33</th>\n",
       "      <td>50.0</td>\n",
       "      <td>0.010000</td>\n",
       "      <td>40.000000</td>\n",
       "      <td>0.844001</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>34</th>\n",
       "      <td>50.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>40.000000</td>\n",
       "      <td>0.843685</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35</th>\n",
       "      <td>50.0</td>\n",
       "      <td>0.000100</td>\n",
       "      <td>40.000000</td>\n",
       "      <td>0.843239</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    factors  regularization      alpha       auc\n",
       "0      25.0        0.000100   1.000000  0.870082\n",
       "1      25.0        1.000000   1.000000  0.868217\n",
       "2      10.0        1.000000   1.000000  0.867958\n",
       "3      10.0        0.000100   1.000000  0.867649\n",
       "4      10.0        0.010000   1.000000  0.867570\n",
       "5      10.0        1.000000  15.000000  0.867010\n",
       "6      25.0        0.010000   1.000000  0.866539\n",
       "7      10.0        0.010000  15.000000  0.866039\n",
       "8      10.0        0.000100  40.000000  0.865394\n",
       "9      10.0        0.000100  15.000000  0.865114\n",
       "10     10.0        0.010000  30.000000  0.865035\n",
       "11     10.0        1.000000  40.000000  0.864732\n",
       "12     10.0        0.000100  30.000000  0.864139\n",
       "13     10.0        0.010000  40.000000  0.863873\n",
       "14     10.0        1.000000  30.000000  0.863563\n",
       "45     19.0        0.574960   5.197704  0.862357\n",
       "15     25.0        0.010000  15.000000  0.861417\n",
       "16     50.0        0.000100   1.000000  0.861343\n",
       "46     23.0        0.562831   8.646115  0.861057\n",
       "17     50.0        0.010000   1.000000  0.860394\n",
       "38     24.0        0.381937  23.055280  0.859475\n",
       "18     50.0        1.000000   1.000000  0.859469\n",
       "19     25.0        1.000000  15.000000  0.859159\n",
       "20     25.0        0.000100  15.000000  0.858931\n",
       "39     32.0        0.777618  10.383273  0.858479\n",
       "44     22.0        0.635830  23.529467  0.858194\n",
       "21     25.0        0.010000  30.000000  0.856522\n",
       "36     47.0        0.421197   3.325613  0.856269\n",
       "22     25.0        0.000100  30.000000  0.856226\n",
       "23     25.0        1.000000  30.000000  0.855834\n",
       "43     31.0        0.588705  17.116074  0.855649\n",
       "24     25.0        0.000100  40.000000  0.855512\n",
       "40     30.0        0.104626  30.460253  0.854976\n",
       "25     25.0        1.000000  40.000000  0.854681\n",
       "26     25.0        0.010000  40.000000  0.854354\n",
       "47     28.0        0.573209  35.403291  0.853781\n",
       "42     40.0        0.053052  40.129906  0.851644\n",
       "37     31.0        0.013218  38.662323  0.850583\n",
       "27     50.0        1.000000  15.000000  0.849484\n",
       "28     50.0        0.010000  15.000000  0.848292\n",
       "41     48.0        0.818650  33.264132  0.847414\n",
       "29     50.0        0.000100  30.000000  0.846366\n",
       "30     50.0        0.000100  15.000000  0.845909\n",
       "31     50.0        1.000000  30.000000  0.845448\n",
       "32     50.0        0.010000  30.000000  0.844548\n",
       "33     50.0        0.010000  40.000000  0.844001\n",
       "34     50.0        1.000000  40.000000  0.843685\n",
       "35     50.0        0.000100  40.000000  0.843239"
      ]
     },
     "execution_count": 42,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "resultsDF = resultsDF.sort_values('auc', ascending = False)\n",
    "resultsDF"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "#### Question 10.2 \n",
    "<div class=\"alert alert-info\">\n",
    "Using \"optimal\" hyper-parameters in question 10.1, re-train the model and show top-5 artist names recommendated for user `2093760`.\n",
    "</div>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<div class=\"alert alert-success\" style=\"background-color: AliceBlue; border-color: CornflowerBlue \"> \n",
    "\n",
    "After testing the model's parameters for achieving the best perforamance represented the the AUC, we obtained the maximum at <b>0.870082</b> for the parameters<br>\n",
    "factors = 25<br>\n",
    "regularization = 0.0001<br>\n",
    "alpha = 1<br>\n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 112,
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "ename": "NameError",
     "evalue": "name 'train_sparse_item_user' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-112-a2b3684fab8b>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m      2\u001b[0m \u001b[0mmodel\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mimplicit\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mals\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mAlternatingLeastSquares\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mfactors\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;36m25\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mregularization\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m0.0001\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0miterations\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m5\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      3\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 4\u001b[0;31m \u001b[0mtrain_data_conf\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m(\u001b[0m\u001b[0mtrain_sparse_item_user\u001b[0m \u001b[0;34m*\u001b[0m \u001b[0;36m1\u001b[0m \u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mastype\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'double'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      5\u001b[0m \u001b[0mmodel\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mfit\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtrain_data_conf\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      6\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mNameError\u001b[0m: name 'train_sparse_item_user' is not defined"
     ]
    }
   ],
   "source": [
    "\n",
    "model = implicit.als.AlternatingLeastSquares(factors = 25, regularization=0.0001, iterations=5)\n",
    "\n",
    "train_data_conf = (train_sparse_item_user * 1 ).astype('double')\n",
    "model.fit(train_data_conf)\n",
    "\n",
    "user_id = 2093760\n",
    "top = 5\n",
    "for incremental_artist_id, score in model.recommend(userid = get_incrementalUserID(user_id), user_items = train_sparse_user_item, N =5):\n",
    "    print(\"Top\",top)\n",
    "    top -= 1\n",
    "    print(\"Artist Name:\",get_artist_name(get_artistID(incremental_artist_id)))\n",
    "    print(\"Score:\",score)\n",
    "    print(\"===========================\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 115,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x864 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#Create tuples (number of play counts, confidence of the 1st recomendation) for 149 users ranging from top to bottom play count\n",
    "'''\n",
    "The top ranked users were removed\n",
    "for the sake of making the plot more clear\n",
    "as their confidence are way higher that\n",
    "they ruin the confidence scale\n",
    "'''\n",
    "playCount_Confidence = [(rankedUsers.iloc[i]['playCount'], model.recommend(userid = get_incrementalUserID(userPlayCount.iloc[i]['userID']), user_items = sparse_user_item, N =1)[0][1])  for i in range(0, userPlayCount.count().values[0]-1, 1000)]\n",
    "\n",
    "\n",
    "\n",
    "#Plot PlayCounts and accuracy of the recommendation per user\n",
    "\n",
    "fig = plt.figure(1)\n",
    "\n",
    "\n",
    "ax1 = plt.subplot(2, 1, 1)\n",
    "\n",
    "ax1.semilogx()\n",
    "plt.scatter(*zip(*playCount_Confidence))\n",
    "ax1.set_xlabel('Number of Plays')\n",
    "ax1.set_ylabel('Recommendation Confidence')\n",
    "ax1.set_title('Relation between Play Count and Confidence in the Recommendation per User')\n",
    "\n",
    "\n",
    "#Create tuples (number of diferent artist the user has search, confidence of the 1st recomendation) for the 149 users ranging from top to bottom curiosity\n",
    "curiosity_Confidence = [(rankedUsersCuriosity.iloc[i]['realArtistID'], model.recommend(userid = get_incrementalUserID(rankedUsersCuriosity.iloc[i]['userID']), user_items = sparse_user_item, N =1)[0][1])  for i in range(0, userPlayCount.count().values[0]-1, 1000)]\n",
    "\n",
    "\n",
    "\n",
    "ax2 = plt.subplot(2, 1, 2)\n",
    "\n",
    "ax2.semilogx()\n",
    "plt.scatter(*zip(*curiosity_Confidence))\n",
    "ax2.set_xlabel('Number of Artists a User has Searched')\n",
    "ax2.set_ylabel('Recommendation Confidence')\n",
    "ax2.set_title('Relation between the Curiosity and Confidence in the Recommendation per User')\n",
    "\n",
    "fig.set_size_inches(w=12,h=12)\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<div class=\"alert alert-success\" style=\"background-color: AliceBlue; border-color: CornflowerBlue \"> \n",
    "After tuning the parameters and training the model again, we've tried to re-observe the relation between the recommendation confidence with each of the user curiosity and the play count per user. However, we achieved more or less the same shapes as in the last case. The key difference was that after tuning the parameters, a higher general confidence was achieved for most of the users in both plots.\n",
    "</div>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "# Summary\n",
    "\n",
    "In this notebook, we introduce an algorithm to do matrix factorization and the way of using it to make recommendation. Further more, we studied how to build a large-scale recommender system using ALS algorithm and evaluate its quality. Finally, a simple approach to choose good parameters is mentioned."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# References\n",
    "\n",
    "- The example in section 2 is taken from [Recommender system](infolab.stanford.edu/~ullman/mmds/ch9.pdf)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.5.2"
  },
  "rise": {
   "scroll": true,
   "theme": "white",
   "transition": "zoom"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}