STEP7_ClusteringEnronKaggle.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "execution": {
     "iopub.execute_input": "2021-11-27T18:44:45.921964Z",
     "iopub.status.busy": "2021-11-27T18:44:45.921084Z",
     "iopub.status.idle": "2021-11-27T18:44:45.926250Z",
     "shell.execute_reply": "2021-11-27T18:44:45.925370Z",
     "shell.execute_reply.started": "2021-11-27T18:44:45.921921Z"
    }
   },
   "outputs": [],
   "source": [
    "import numpy as np # linear algebra\n",
    "import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)\n",
    "import os"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-11-27T18:44:45.932252Z",
     "iopub.status.busy": "2021-11-27T18:44:45.931651Z",
     "iopub.status.idle": "2021-11-27T18:44:45.957623Z",
     "shell.execute_reply": "2021-11-27T18:44:45.956641Z",
     "shell.execute_reply.started": "2021-11-27T18:44:45.932211Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[]\n",
      "[]\n",
      "['gcp.py']\n",
      "../lib/kaggle/gcp.py\n",
      "[]\n",
      "['enron.csv']\n",
      "../input/enronemails/enron.csv\n",
      "['Enron_29_Features.csv']\n",
      "../input/enron29features/Enron_29_Features.csv\n",
      "['__notebook_source__.ipynb']\n",
      "../working/__notebook_source__.ipynb\n"
     ]
    }
   ],
   "source": [
    "for dirname, _, filenames in os.walk('../'):\n",
    "    print(filenames)\n",
    "    for filename in filenames:\n",
    "        print(os.path.join(dirname, filename))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-11-27T18:44:45.959376Z",
     "iopub.status.busy": "2021-11-27T18:44:45.959148Z",
     "iopub.status.idle": "2021-11-27T18:44:46.953277Z",
     "shell.execute_reply": "2021-11-27T18:44:46.952665Z",
     "shell.execute_reply.started": "2021-11-27T18:44:45.959349Z"
    }
   },
   "outputs": [],
   "source": [
    "df = pd.read_csv(\"../input/enron29features/Enron_29_Features.csv\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This notebook was actually run on kaggle, as K-means clustering was not working on our local machines. Here we are using the Enron 29 Feature Dataset that we created extracting all the stylometric features from all the author emails found in the initial dataset Enron.csv"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-11-27T19:07:14.554894Z",
     "iopub.status.busy": "2021-11-27T19:07:14.553758Z",
     "iopub.status.idle": "2021-11-27T19:07:34.601817Z",
     "shell.execute_reply": "2021-11-27T19:07:34.600790Z",
     "shell.execute_reply.started": "2021-11-27T19:07:14.554837Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[nltk_data] Error loading stopwords: <urlopen error [Errno -3]\n",
      "[nltk_data]     Temporary failure in name resolution>\n"
     ]
    }
   ],
   "source": [
    "import re\n",
    "import string\n",
    "from sklearn.preprocessing import StandardScaler, LabelEncoder\n",
    "import nltk\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "from sklearn.cluster import KMeans\n",
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "from collections import Counter \n",
    "from sklearn.linear_model import LogisticRegression\n",
    "from gensim.models import Word2Vec\n",
    "\n",
    "from nltk import word_tokenize\n",
    "from nltk.corpus import stopwords\n",
    "\n",
    "from sklearn.cluster import MiniBatchKMeans\n",
    "from sklearn.metrics import silhouette_samples, silhouette_score\n",
    "from sklearn.metrics.pairwise import euclidean_distances\n",
    "nltk.download(\"stopwords\")\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "from nltk.tokenize import word_tokenize\n",
    "from nltk.stem import WordNetLemmatizer\n",
    "from nltk.stem import PorterStemmer\n",
    "from sklearn import preprocessing\n",
    "from tqdm import tqdm\n",
    "from keras.models import Sequential\n",
    "from keras.layers.recurrent import LSTM, GRU\n",
    "from keras.layers.core import Dense, Activation, Dropout\n",
    "from keras.layers.embeddings import Embedding\n",
    "from tensorflow.keras.layers import BatchNormalization\n",
    "from keras.utils import np_utils\n",
    "from keras.layers import GlobalMaxPooling1D, Conv1D, MaxPooling1D, Flatten, Bidirectional, SpatialDropout1D\n",
    "from sklearn.model_selection import train_test_split\n",
    "from keras.preprocessing import sequence\n",
    "from keras.preprocessing.text import Tokenizer\n",
    "from keras.callbacks import EarlyStopping\n",
    "from sklearn.model_selection import KFold\n",
    "import matplotlib.pyplot as plt\n",
    "import tensorflow as tf\n",
    "import string\n",
    "import re\n",
    "from collections import Counter"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-11-27T19:07:34.603577Z",
     "iopub.status.busy": "2021-11-27T19:07:34.603374Z",
     "iopub.status.idle": "2021-11-27T19:07:34.608750Z",
     "shell.execute_reply": "2021-11-27T19:07:34.607702Z",
     "shell.execute_reply.started": "2021-11-27T19:07:34.603552Z"
    }
   },
   "outputs": [],
   "source": [
    "from nltk.corpus import stopwords\n",
    "stop_words = stopwords.words(\"english\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-11-27T19:07:34.610520Z",
     "iopub.status.busy": "2021-11-27T19:07:34.610284Z",
     "iopub.status.idle": "2021-11-27T19:07:34.627032Z",
     "shell.execute_reply": "2021-11-27T19:07:34.626169Z",
     "shell.execute_reply.started": "2021-11-27T19:07:34.610492Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "mann-k          4000\n",
       "kaminski-v      4000\n",
       "dasovich-j      4000\n",
       "germany-c       4000\n",
       "shackleton-s    4000\n",
       "Name: Folder, dtype: int64"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df[\"Folder\"].value_counts()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Here we are testing with 5 authors, 4000 emails per author"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-11-27T19:07:34.629245Z",
     "iopub.status.busy": "2021-11-27T19:07:34.629024Z",
     "iopub.status.idle": "2021-11-27T19:07:34.638792Z",
     "shell.execute_reply": "2021-11-27T19:07:34.637789Z",
     "shell.execute_reply.started": "2021-11-27T19:07:34.629220Z"
    }
   },
   "outputs": [],
   "source": [
    "def text_process(text):\n",
    "    #Remove Punctuation Marks\n",
    "    text = text.lower()\n",
    "    nopunct = \"\"\n",
    "    clean_final = []\n",
    "    for char in text:\n",
    "        if re.match(r\"\\w\", char) or re.match(r\" \", char):\n",
    "            nopunct += char\n",
    "        elif re.match(r\" \", char):\n",
    "            nopunct += char\n",
    "        else:\n",
    "            nopunct += \" \"\n",
    "    for word in nopunct.split():\n",
    "        if not word in stop_words:\n",
    "            clean_final.append(word)\n",
    "    nopunct=\" \".join(clean_final)\n",
    "    return nopunct"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-11-27T19:07:34.640659Z",
     "iopub.status.busy": "2021-11-27T19:07:34.640409Z",
     "iopub.status.idle": "2021-11-27T19:07:41.163535Z",
     "shell.execute_reply": "2021-11-27T19:07:41.162585Z",
     "shell.execute_reply.started": "2021-11-27T19:07:34.640632Z"
    }
   },
   "outputs": [],
   "source": [
    "df = df[df[\"Email Length\"].notna()]\n",
    "df = df[df[\"Author\"].notna()]\n",
    "clean_text = df[\"Text\"].apply(lambda row: text_process(row))\n",
    "df[\"Processed Text\"] = clean_text"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Here we performed text preprocessing, removing all the stop words, and non alphanumeric characters"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-11-27T19:15:41.932799Z",
     "iopub.status.busy": "2021-11-27T19:15:41.932466Z",
     "iopub.status.idle": "2021-11-27T19:15:41.939307Z",
     "shell.execute_reply": "2021-11-27T19:15:41.938491Z",
     "shell.execute_reply.started": "2021-11-27T19:15:41.932764Z"
    }
   },
   "outputs": [],
   "source": [
    "df[\"Processed Text\"] = proccessed_text"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-11-27T19:15:45.515787Z",
     "iopub.status.busy": "2021-11-27T19:15:45.515498Z",
     "iopub.status.idle": "2021-11-27T19:15:48.521748Z",
     "shell.execute_reply": "2021-11-27T19:15:48.520955Z",
     "shell.execute_reply.started": "2021-11-27T19:15:45.515756Z"
    }
   },
   "outputs": [],
   "source": [
    "tokenized_text = df[\"Processed Text\"].apply(lambda row: word_tokenize(row))\n",
    "df[\"Tokens\"] = tokenized_text"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-11-27T19:07:44.276800Z",
     "iopub.status.busy": "2021-11-27T19:07:44.276566Z",
     "iopub.status.idle": "2021-11-27T19:07:44.282619Z",
     "shell.execute_reply": "2021-11-27T19:07:44.282073Z",
     "shell.execute_reply.started": "2021-11-27T19:07:44.276766Z"
    }
   },
   "outputs": [],
   "source": [
    "def vectorize(list_of_docs, model):\n",
    "\n",
    "    features = []\n",
    "\n",
    "    for tokens in list_of_docs:\n",
    "        zero_vector = np.zeros(model.vector_size)\n",
    "        vectors = []\n",
    "        for token in tokens:\n",
    "            if token in model.wv:\n",
    "                try:\n",
    "                    vectors.append(model.wv[token])\n",
    "                except KeyError:\n",
    "                    continue\n",
    "        if vectors:\n",
    "            vectors = np.asarray(vectors)\n",
    "            avg_vec = vectors.mean(axis=0)\n",
    "            features.append(avg_vec)\n",
    "        else:\n",
    "            features.append(zero_vector)\n",
    "    return features"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This function takes input of a list of tokens, a trained word2vec model. The function then goes through all the tokens and finds the appropriate word2vec vector and appends the found vector to a list called vectors. To represent the all the word vectors present in the list vectors, we take the average of the vectors and return that to represent the list of tokens passed into the function. If none of the tokens passed in were found in the glove word dictionary a zero vector is returned."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-11-27T19:07:44.284027Z",
     "iopub.status.busy": "2021-11-27T19:07:44.283779Z",
     "iopub.status.idle": "2021-11-27T19:07:44.298508Z",
     "shell.execute_reply": "2021-11-27T19:07:44.297868Z",
     "shell.execute_reply.started": "2021-11-27T19:07:44.283998Z"
    }
   },
   "outputs": [],
   "source": [
    "X = df[\"Tokens\"].to_numpy()\n",
    "y = df[\"Folder\"]\n",
    "labelencoder = LabelEncoder()\n",
    "y = labelencoder.fit_transform(y)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Here we are label encoding our target variable Y, which in this case is the folder name or author name of the sent email."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-11-27T19:07:44.301672Z",
     "iopub.status.busy": "2021-11-27T19:07:44.301162Z",
     "iopub.status.idle": "2021-11-27T19:07:48.791364Z",
     "shell.execute_reply": "2021-11-27T19:07:48.790770Z",
     "shell.execute_reply.started": "2021-11-27T19:07:44.301632Z"
    }
   },
   "outputs": [],
   "source": [
    "X_train, X_test, y_train, y_test = train_test_split(df, y, test_size=0.33)\n",
    "\n",
    "tokenized_docs = X_train[\"Tokens\"]\n",
    "model = Word2Vec(sentences=tokenized_docs, vector_size=100, workers=1, seed=42)\n",
    "vectorized_docs = vectorize(tokenized_docs, model=model)\n",
    "\n",
    "test_tokenized_docs = X_test[\"Tokens\"]\n",
    "test_vectorized_docs = vectorize(test_tokenized_docs, model=model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-11-27T19:07:48.794247Z",
     "iopub.status.busy": "2021-11-27T19:07:48.793650Z",
     "iopub.status.idle": "2021-11-27T19:07:48.820386Z",
     "shell.execute_reply": "2021-11-27T19:07:48.819841Z",
     "shell.execute_reply.started": "2021-11-27T19:07:48.794202Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(13400, 33)\n",
      "(13400, 100)\n",
      "(6600, 33)\n",
      "(6600, 100)\n"
     ]
    }
   ],
   "source": [
    "print(X_train.shape)\n",
    "vect_np = np.asarray(vectorized_docs)\n",
    "print(vect_np.shape)\n",
    "print(X_test.shape)\n",
    "vect_test_np = np.asarray(test_vectorized_docs)\n",
    "print(vect_test_np.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-11-27T19:07:48.821686Z",
     "iopub.status.busy": "2021-11-27T19:07:48.821212Z",
     "iopub.status.idle": "2021-11-27T19:07:48.827229Z",
     "shell.execute_reply": "2021-11-27T19:07:48.826543Z",
     "shell.execute_reply.started": "2021-11-27T19:07:48.821658Z"
    }
   },
   "outputs": [],
   "source": [
    "def calculate_WSS(points, kmax):\n",
    "    sse = []\n",
    "    for k in range(1, kmax+1):\n",
    "        kmeans = KMeans(n_clusters = k).fit(points)\n",
    "        centroids = kmeans.cluster_centers_\n",
    "        pred_clusters = kmeans.predict(points)\n",
    "        curr_sse = 0\n",
    "    \n",
    "    # calculate square of Euclidean distance of each point from its cluster center and add to current WSS\n",
    "        for i in range(len(points)):\n",
    "            curr_center = centroids[pred_clusters[i]]\n",
    "            ed = euclidean_distances(points[i].reshape(1, -1), curr_center.reshape(1, -1))\n",
    "            curr_sse += ed\n",
    "        sse.append(curr_sse)\n",
    "    return sse"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-11-27T19:07:48.828526Z",
     "iopub.status.busy": "2021-11-27T19:07:48.828236Z",
     "iopub.status.idle": "2021-11-27T19:10:37.249612Z",
     "shell.execute_reply": "2021-11-27T19:10:37.248716Z",
     "shell.execute_reply.started": "2021-11-27T19:07:48.828502Z"
    }
   },
   "outputs": [],
   "source": [
    "wss = calculate_WSS(vect_np, 30)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Here we are using WSS, to try to find the optimal number of K"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-11-27T19:19:39.220647Z",
     "iopub.status.busy": "2021-11-27T19:19:39.219822Z",
     "iopub.status.idle": "2021-11-27T19:19:39.419159Z",
     "shell.execute_reply": "2021-11-27T19:19:39.418206Z",
     "shell.execute_reply.started": "2021-11-27T19:19:39.220605Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[array([[25705.48926217]]), array([[22482.99491653]]), array([[20079.88305877]]), array([[18792.95656652]]), array([[17686.85691677]]), array([[16980.02833806]]), array([[16477.73525058]]), array([[16006.17961776]]), array([[15553.66548428]]), array([[15115.00972873]]), array([[14814.09882667]]), array([[14500.04011358]]), array([[14224.98211133]]), array([[14154.64197252]]), array([[13950.52470367]]), array([[13722.86383607]]), array([[13484.08703379]]), array([[13348.44935733]]), array([[13200.31707173]]), array([[13171.76303373]]), array([[12903.79969465]]), array([[12758.62243523]]), array([[12604.69778695]]), array([[12560.73533029]]), array([[12345.5356474]]), array([[12222.04736653]]), array([[12123.30823895]]), array([[12035.42471798]]), array([[11917.67434602]]), array([[11776.01771248]])]\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYMAAAEGCAYAAACHGfl5AAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjQuMywgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/MnkTPAAAACXBIWXMAAAsTAAALEwEAmpwYAAAsIklEQVR4nO3deXxV9Z3/8dcnGwTISkIgCRBkD6AIUam7VBCsHbTbaJ2KU6fOdLTV1jrWdqa2nbaj1Wrr1NqfWyutlVqXSjsqYotbK0JAWQICEUESlkS2gGxZPr8/zgleYyAh2829eT8fj/vIvZ9zzr3f7+NC3jnne873mLsjIiI9W0K0GyAiItGnMBAREYWBiIgoDEREBIWBiIgASdFuQFvl5OR4UVFRtJshIhJTli5d+p675zatx2wYFBUVUVpaGu1miIjEFDPb1Fxdh4lERKTlMDCzwWa20MxWm1mZmV0XsewrZvZWWP9xRP1mMys3s7VmdkFEfUZYKzezb0bUh5nZ62H992aW0pGdFBGRY2vNYaI64AZ3X2ZmacBSM1sA5AGzgJPc/ZCZDQAws2LgUmAckA+8YGajwve6B5gGVABLzGyeu68GbgPucve5ZvZL4Crg3o7rpoiIHEuLewbuvtXdl4XP9wJrgALgy8Ct7n4oXFYVbjILmOvuh9z9HaAcODV8lLv7Bnc/DMwFZpmZAVOBx8PtHwYu7qD+iYhIKxzXmIGZFQEnA68Do4CzwsM7L5nZKeFqBcDmiM0qwtrR6v2B3e5e16Te3OdfbWalZlZaXV19PE0XEZFjaHUYmFk/4AngenevITjElA1MAW4EHgv/yu807n6fu5e4e0lu7kfOjBIRkTZq1amlZpZMEASPuPuTYbkCeNKDaU8Xm1kDkANUAoMjNi8MaxylvgPINLOkcO8gcn0REekCrTmbyIAHgTXufmfEoj8C54XrjAJSgPeAecClZtbLzIYBI4HFwBJgZHjmUArBIPO8MEwWAp8J33c28HQH9K1Zv3ltI39avqWz3l5EJCa1Zs/gDOALwEozezOsfQt4CHjIzFYBh4HZ4S/2MjN7DFhNcCbSNe5eD2Bm1wLzgUTgIXcvC9/vJmCumf0AeIMgfDrFY6UV9O2VyCdPyu+sjxARiTkthoG7vwocbSzgn46yzQ+BHzZTfwZ4ppn6BoKzjTrd5KFZzF3yLrX1DSQn6po7ERHogVcglxRlcbC2gTVba6LdFBGRbqPHhcHkoVkAlG7cFeWWiIh0Hz0uDAZlpFKQmcrSTQoDEZFGPS4MINg7KN20k2C8W0REemQYlBRlsb3mEJW7D0S7KSIi3UKPDIPGcQMdKhIRCfTIMBidl0bflEQNIouIhHpkGCQlJnDykCxKtWcgIgL00DCA4FDR2m017D1YG+2miIhEXY8Ng5KiLBoc3ty8O9pNERGJuh4bBhMHZ5JguvhMRAR6cBik9U5m9MB0lr2rMBAR6bFhAFAyNIs33t1NfYMuPhORnq1nh0FRFvsO1fHWNk1aJyI9W48Og0lDdPGZiAj08DAozEolL72XwkBEerweHQZmRsnQbJ1RJCI9Xo8OA4BJQ7Oo3H2AbXsORrspIiJR0+PDoKTxZjebdka5JSIi0dNiGJjZYDNbaGarzazMzK5rsvwGM3Mzywlfm5ndbWblZrbCzCZFrDvbzNaHj9kR9clmtjLc5m4zO9o9lztccX46qcmJGjcQkR6tNXsGdcAN7l4MTAGuMbNiCIICmA68G7H+TGBk+LgauDdcNxu4BTgNOBW4xcyywm3uBb4Usd2M9nWr9ZITEzhpcIbCQER6tBbDwN23uvuy8PleYA1QEC6+C/gPIPKqrVnAHA8sAjLNbBBwAbDA3Xe6+y5gATAjXJbu7os8uPXYHODijule60wemkXZlhr2H67ryo8VEek2jmvMwMyKgJOB181sFlDp7subrFYAbI54XRHWjlWvaKbe3OdfbWalZlZaXV19PE0/ppKh2dQ3uCatE5Eeq9VhYGb9gCeA6wkOHX0L+E7nNKt57n6fu5e4e0lubm6HvW/jxWfLdKhIRHqoVoWBmSUTBMEj7v4kMBwYBiw3s41AIbDMzAYClcDgiM0Lw9qx6oXN1LtMRp9kRuX1081uRKTHas3ZRAY8CKxx9zsB3H2luw9w9yJ3LyI4tDPJ3bcB84ArwrOKpgB73H0rMB+YbmZZ4cDxdGB+uKzGzKaEn3UF8HQn9PWYJg/NYtmmXTRo0joR6YFas2dwBvAFYKqZvRk+LjzG+s8AG4By4H7g3wHcfSfw38CS8PH9sEa4zgPhNm8Dz7ahL+0yeWg2NQfrWF+1r6s/WkQk6pJaWsHdXwWOed5/uHfQ+NyBa46y3kPAQ83US4HxLbWlMzVefLZ00y5GD0yLZlNERLpcj78CudHQ/n3I6ZeiK5FFpEdSGITMjElDsnTxmYj0SAqDCCVFWWzasZ/qvYei3RQRkS6lMIgweWg2oJvdiEjPozCIML4gnZSkBJZq3EBEehiFQYReSYmcWJChi89EpMdRGDQxuSiLVZV7OFhbH+2miIh0GYVBEyVDs6mtd1ZW7ol2U0REuozCoInJjXc+032RRaQHURg0kd03hRNy+moQWUR6FIVBMyYPDS4+C2bWEBGJfwqDZpQUZbFrfy0b3ns/2k0REekSCoNmHLn4TOMGItJDKAyacUJOXzL7JGvSOhHpMRQGzUhIMCYPydLFZyLSYygMjmJyURYbqt9n5/uHo90UEZFOpzA4itOH5wDw0rqqKLdERKTzKQyO4sSCDAak9eKF1QoDEYl/CoOjSEgwPj42jxfXVnGoTvMUiUh8Uxgcw/TiPN4/XM9rb++IdlNERDpVi2FgZoPNbKGZrTazMjO7LqzfbmZvmdkKM3vKzDIjtrnZzMrNbK2ZXRBRnxHWys3smxH1YWb2elj/vZmldHA/2+Rjw/vTJyWRBau3R7spIiKdqjV7BnXADe5eDEwBrjGzYmABMN7dTwTWATcDhMsuBcYBM4BfmFmimSUC9wAzgWLgsnBdgNuAu9x9BLALuKqjOtgevZMTOXtkLi+s2U5Dg6amEJH41WIYuPtWd18WPt8LrAEK3P15d68LV1sEFIbPZwFz3f2Qu78DlAOnho9yd9/g7oeBucAsMzNgKvB4uP3DwMUd0rsOMK04j+01h1i1RVNai0j8Oq4xAzMrAk4GXm+y6IvAs+HzAmBzxLKKsHa0en9gd0SwNNab+/yrzazUzEqrq6uPp+ltNnXMABITTIeKRCSutToMzKwf8ARwvbvXRNS/TXAo6ZGOb96Huft97l7i7iW5ubmd/XEAZPVNoWRolsJAROJaq8LAzJIJguARd38yon4lcBFwuX8w33MlMDhi88KwdrT6DiDTzJKa1LuNacV5vLVtL5t37o92U0REOkVrziYy4EFgjbvfGVGfAfwH8A/uHvlbch5wqZn1MrNhwEhgMbAEGBmeOZRCMMg8LwyRhcBnwu1nA0+3v2sdZ1pxHgDPa+9AROJUa/YMzgC+AEw1szfDx4XAz4E0YEFY+yWAu5cBjwGrgeeAa9y9PhwTuBaYTzAI/Vi4LsBNwNfNrJxgDOHBjuti+w3t35dRef14QWEgInEqqaUV3P1VwJpZ9Mwxtvkh8MNm6s80t527byA426jbmlacxy9f2sDu/YfJ7NMtLoMQEekwugK5laYVD6S+wVm4VnMViUj8URi0UuPEdTqrSETikcKglRonrntpbbUmrhORuKMwOA6auE5E4pXC4Dho4joRiVcKg+OgietEJF4pDI5T48R1Kys1cZ2IxA+FwXHSxHUiEo8UBsepceK6F9YoDEQkfigM2kAT14lIvFEYtIEmrhOReKMwaIPGiesWrN4W7aaIiHQIhUEbTSvOY8nGXezefzjaTRERaTeFQRtp4joRiScKgzbSxHUiEk8UBm2kietEJJ4oDNqhceK6v2viOhGJcQqDdtDEdSISLxQG7dA7OZFzRuXyF01cJyIxrsUwMLPBZrbQzFabWZmZXRfWs81sgZmtD39mhXUzs7vNrNzMVpjZpIj3mh2uv97MZkfUJ5vZynCbu82suXsud0uauE5E4kFr9gzqgBvcvRiYAlxjZsXAN4G/uPtI4C/ha4CZwMjwcTVwLwThAdwCnAacCtzSGCDhOl+K2G5G+7vWNc4bHUxcN79MF6CJSOxqMQzcfau7Lwuf7wXWAAXALODhcLWHgYvD57OAOR5YBGSa2SDgAmCBu+90913AAmBGuCzd3Re5uwNzIt6r28vqm8J5o3P5zaJN7Nh3KNrNERFpk+MaMzCzIuBk4HUgz923hou2AXnh8wJgc8RmFWHtWPWKZurNff7VZlZqZqXV1dXH0/RO9c2ZY9h/uJ47F6yLdlNERNqk1WFgZv2AJ4Dr3b0mcln4F32nj6C6+33uXuLuJbm5uZ39ca02YkAaX5gylEcXv8uarTUtbyAi0s20KgzMLJkgCB5x9yfD8vbwEA/hz8Z5GSqBwRGbF4a1Y9ULm6nHlOvPH0l6ajLf/9NqgmwUEYkdrTmbyIAHgTXufmfEonlA4xlBs4GnI+pXhGcVTQH2hIeT5gPTzSwrHDieDswPl9WY2ZTws66IeK+Ykdknha9PG8VrG3Ywv0zXHYhIbGnNnsEZwBeAqWb2Zvi4ELgVmGZm64Hzw9cAzwAbgHLgfuDfAdx9J/DfwJLw8f2wRrjOA+E2bwPPdkDfutznTx3CqLx+/OiZNZqiQkRiisXqIY2SkhIvLS2NdjM+4pX11XzhwcXcNGMMXz53eLSbIyLyIWa21N1LmtZ1BXIHO2tkLuePHcDP/7qeqr0Ho90cEZFWURh0gm9/opjD9Q3cMX9ttJsiItIqCoNOMCynL/98xjD+sLSClRWapkJEuj+FQSe5duoIsvuk8L0/lelUUxHp9hQGnSS9dzLfuGA0pZt28ecVW1veQEQkihQGnehzJYMZOyidW599i4O1OtVURLovhUEnSkwwbvlkMZW7D3Dfyxui3RwRkaNSGHSyKSf0Z+b4gdz74tts26NTTUWke1IYdIFvXTiWendue+6taDdFRKRZCoMuMDi7D/9y5jCeeqOSZe/uinZzREQ+QmHQRf79vBHkpvXie39arfsli0i3ozDoIv16JXHTjDEs37ybR17fFO3miIh8iMKgC316UgFnjczhf559i80790e7OSIiRygMupCZceunTyTBjBsfX67DRSLSbSgMulhBZir/ddFYFm3YyW91uEhEugmFQRR8rmQw54zK5X+eeYt3d+hwkYhEn8IgCoLDRRNISjC+ocNFItINKAyiZFBGKv91UTGL39nJnNc2Rrs5ItLDKQyi6LMlhZw7OpfbnlvLph3vR7s5ItKDtRgGZvaQmVWZ2aqI2kQzW2Rmb5pZqZmdGtbNzO42s3IzW2FmkyK2mW1m68PH7Ij6ZDNbGW5zt5lZR3eyuzIz/udTE0hKNG58fIUOF4lI1LRmz+DXwIwmtR8D33P3icB3wtcAM4GR4eNq4F4AM8sGbgFOA04FbjGzrHCbe4EvRWzX9LPi2qCMVL4THi56WIeLRCRKWgwDd38Z2Nm0DKSHzzOALeHzWcAcDywCMs1sEHABsMDdd7r7LmABMCNclu7uizy4Hdgc4OL2dirWfGZyIVPHDOC2595i43s6XCQiXa+tYwbXA7eb2WbgDuDmsF4AbI5YryKsHate0Uy9RzEzfnTJBJITE3QxmohERVvD4MvA19x9MPA14MGOa9LRmdnV4RhFaXV1dVd8ZJcZmNGbWz45jiUbd/Grv2+MdnNEpIdpaxjMBp4Mn/+BYBwAoBIYHLFeYVg7Vr2wmXqz3P0+dy9x95Lc3Nw2Nr37+vSkAj4+ZgC3z3+Ld3S4SES6UFvDYAtwTvh8KrA+fD4PuCI8q2gKsMfdtwLzgelmlhUOHE8H5ofLasxsSngW0RXA023tTKwzM370qQmkJCZw4x+WU6/DRSLSRVpzaumjwGvAaDOrMLOrCM7++YmZLQd+RHDmEMAzwAagHLgf+HcAd98J/DewJHx8P6wRrvNAuM3bwLMd07XYlJfem+/+wzhKN+3i3hfLo90cEekhLDiJJ/aUlJR4aWlptJvRKdyd6+a+ybzlW7jxgtFcc96IaDdJROKEmS1195Km9aRoNEaOzcy483MnkWBw+/y17D9cxzemj6YHXY8nIl1MYdBNJSUm8JPPTSQ1JZF7Fr7N/sP1fOeiYgWCiHQKhUE3lpgQXH/QOzmRX/1tIwdr6/nBxRNITFAgiEjHUhh0c2bGdy4qpm9KEj9fWM6Bw/Xc8dmTSErUHIMi0nEUBjHAzPjGBaNJTUnk9vlrOVBbz92XnUyvpMRoN01E4oT+vIwh15w3gu9cVMz8su1cPWcpB2vro90kEYkTCoMY88Uzh3Hrpybw8vpqrvzVYvYdqot2k0QkDigMYtClpw7hp/84kSUbd/GFB19nz/7aaDdJRGKcwiBGzZpYwD2fn8Sqyj1cev8i3SlNRNpFYRDDZowfyP1XlFCxaz8zfvoKv1m0iVi9olxEokthEOPOHT2A+defTUlRFv/1x1Vc8dBituw+EO1miUiMURjEgfzMVOZ88VR+eMl4lm7axQV3vcwfSjdrL0FEWk1hECfMjMtPG8pz153N2EHp3Pj4Cr40p5SqvQej3TQRiQEKgzgzpH8f5l49hf/8xFheXv8e0+96mT+v2NLyhiLSoykM4lBCgvEvZ53AM189k6HZfbj2d29w7e+Wsev9w9Fumoh0UwqDODZiQBpPfPl0brxgNPPLtjHtrpeZX7Yt2s0SkW5IYRDnkhITuOa8ETx9zZnkpvXiX3+zlC//dilVNRpLEJEPKAx6iOL8dOZdewY3XjCav7xVxcfvfIlHF79Lg+6zLCIoDHqU5HAv4bnrzmJcfjo3P7mSS+9fxNvV+6LdNBGJMoVBD3RCbj8e/dIUbvv0BN7aWsPMn73Cz/+6nsN1DdFumohESYthYGYPmVmVma1qUv+Kmb1lZmVm9uOI+s1mVm5ma83sgoj6jLBWbmbfjKgPM7PXw/rvzSylozonR2dm/OMpQ3jhhnOYNjaPO55fxyf/91XeeHdXtJsmIlHQmj2DXwMzIgtmdh4wCzjJ3ccBd4T1YuBSYFy4zS/MLNHMEoF7gJlAMXBZuC7AbcBd7j4C2AVc1d5OSesNSOvNPZdP4oErSqg5WMun7v07351XpqmxRXqYFsPA3V8GdjYpfxm41d0PhetUhfVZwFx3P+Tu7wDlwKnho9zdN7j7YWAuMMuCu7tPBR4Pt38YuLh9XZK2OL84j+e/djZXTBnKw69tZPqdLzG/bJumtBDpIdo6ZjAKOCs8vPOSmZ0S1guAzRHrVYS1o9X7A7vdva5JvVlmdrWZlZpZaXV1dRubLkeT1juZ780az+P/djrpqcn862+W8i8Pl7J55/5oN01EOllbwyAJyAamADcCj4V/5Xcqd7/P3UvcvSQ3N7ezP67Hmjw0iz995Uy+feFYXtuwg2l3vcQ9C8s5VKfbbIrEq7aGQQXwpAcWAw1ADlAJDI5YrzCsHa2+A8g0s6QmdYmy5MQEvnT2Cbzw9XM4b/QAbp+/lpk/e4W/l78X7aaJSCdoaxj8ETgPwMxGASnAe8A84FIz62Vmw4CRwGJgCTAyPHMohWCQeZ4HB6QXAp8J33c28HQb2ySdID8zlXv/aTK/+udTqKt3Pv/A61w/9w3NhioSZ1pzaumjwGvAaDOrMLOrgIeAE8LTTecCs8O9hDLgMWA18BxwjbvXh2MC1wLzgTXAY+G6ADcBXzezcoIxhAc7tovSEc4bPYDnv3Y2X506gmdWbuPjd7zEw3/fSL2uYBaJCxarZ4uUlJR4aWlptJvRI22o3sd/Pb2Kv5XvYHxBOj+4eAITB2dGu1ki0gpmttTdS5rWdQWyHLcTcvvx26tO4+7LTmZ7zSEu+cXf+OYTK9ix71C0myYibaQwkDYxM/7hpHz+esM5/MuZw3h8aQXn3fEic17bSF29prUQiTUKA2mXtN7JfPsTxTx73VlMKMzgO0+XcdH/vsrid5pepygi3ZnCQDrEyLw0fnvVafzi8knUHKjlc//vNa6f+wbbdd8EkZigMJAOY2ZcOGEQL9xwDl8JzzqaeseL3Pfy25oRVaSbUxhIh+uTksQN00fz/NfOZsoJ/fnRM28x82cv89K6as11JNJNKQyk0xTl9OXBK0/hwdkl1DU4sx9azMyfvcJvXtvI3oO10W6eiETQdQbSJQ7W1vPUG5U88vomVlXW0CclkVkT87n8tKGML8iIdvNEeoyjXWegMJAut3zzbh55fRPzlm/hYG0DJxVmcPmUoXzyxHxSUxKj3TyRuKYwkG5nz4FanlpWwSOvv8v6qn2k9U7i05MKufy0IYzMS4t280TiksJAui13Z8nGXfx20SaeXbWV2nrntGHZXHl6EdOK80hK1NCWSEdRGEhMeG/fIf5QWsFvF22icvcB8jN6c/mUoVx6ymD69+sV7eaJxDyFgcSU+gbnhTXbmfPaRv5WvoOUpAQ+eWI+V55exIRCDTiLtNXRwiCpuZVFoi0xwbhg3EAuGDeQ9dv38vBrG3lyWSVPLKvg5CGZXHl6ETPHDyIlSYeQRDqC9gwkZtQcrOXx0grmvLaRjTv2k5vWi8tOHcJnJxcyOLtPtJsnEhN0mEjiRkOD8/L6ah7++0ZeXFeNO5w2LJtPTy7kwgmD6NdLO7wiR6MwkLhUufsATy2r4PGlFWzcsZ/U5ERmjh/IpycX8rET+pOQYNFuoki3ojCQuObuLHt3F48vreTPK7aw92Ad+Rm9uWRSAZ+eVMgJuf2i3USRbkFhID3Gwdp6nl+9nSeWVvDK+moaHCYNyeTUYf3JTetFblovBoQ/c9N6kdYrCTPtQUjP0OYwMLOHgIuAKncf32TZDcAdQK67v2fB/6ifARcC+4Er3X1ZuO5s4D/DTX/g7g+H9cnAr4FU4BngOm9FQikMpDW21xzkj29U8tQblbxdvY/a+o/+0+qdnBAEQ79eDEjrTX5mKh8b3p+PDe+v8QeJO+0Jg7OBfcCcyDAws8HAA8AYYHIYBhcCXyEIg9OAn7n7aWaWDZQCJYADS8NtdpnZYuCrwOsEYXC3uz/bUocUBnK83J09B2qp2nuI6vBRtfdgxPPg5+Zd+zlY20BSgjFpaBbnjMrl7JG5jMtP1xiExLw2X2fg7i+bWVEzi+4C/gN4OqI2iyA0HFhkZplmNgg4F1jg7jvDxiwAZpjZi0C6uy8K63OAi4EWw0DkeJkZmX1SyOyTwqhjzH10qK6epZt28fK693hlfTW3z1/L7fPXkt03hTNH5HD2qFzOHpnDgPTeXdh6kc7Vpn1gM5sFVLr78ibHWguAzRGvK8LaseoVzdSP9rlXA1cDDBkypC1NF2lRr6RETh+ew+nDc/jmzDFU7z3Eq+XVR8Jh3vItAIwZmMZ5YwZw/tg8Th6cqb0GiWnHHQZm1gf4FjC945tzbO5+H3AfBIeJuvrzpWfKTevFJScXcsnJhTQ0OGu21fDK+vd4aW0197+8gXtffJucfr04f+wAphXnccaIHHonaypuiS1t2TMYDgwDGvcKCoFlZnYqUAkMjli3MKxVEhwqiqy/GNYLm1lfpFtKSDDG5WcwLj+DfztnOHsO1PLi2ioWrN7On1dsZe6SzaQmJ3LWyBymFefx8bF5ZPdNiXazRVp03GHg7iuBAY2vzWwjUBIOIM8DrjWzuQQDyHvcfauZzQd+ZGZZ4WbTgZvdfaeZ1ZjZFIIB5CuA/21fl0S6TkZqMrMmFjBrYgGH6xpYtGEHC1Zv54U123l+9XYSDEqGZjOtOI8Z4wdq2gzptlpzNtGjBH/V5wDbgVvc/cGI5Rv5IAwM+Dkwg+DU0n9299JwvS8SHF4C+KG7/yqsl/DBqaXPAl/RqaUS69ydsi01PL96OwtWb2fN1hoAxuWnM2PcQGZOGMiIAbqBj3Q9XXQmEkWbd+7nuVXbeK5sG0s37QJgeG5fZo4fxIzxAxmXn64L36RLKAxEuontNQeZX7aN51ZtY9GGHTQ4DM5OZca4gcwYP4iJgzNJ1JlJ0kkUBiLd0M73D7NgdRAMr5a/R229k5qcyKiBaRQPSmPsoHTGDkpnzMA00nonR7u5EgcUBiLdXM3BWha+VcXyzXtYs7WGNdtq2L2/9sjywdmpjB2YfiQgxuWnU5iVqsNLclx0pzORbi699wdnJkEwCL2t5mAQDFv3snprDWu21rBgzXYa/4bLSE1mQkEG4wsymBA+BmcrIOT4KQxEuikzY1BGKoMyUpk6Ju9I/cDhetZu30vZlj2sqtzDyso9PPjqhiOT8DUXEAMzepOcaAoJOSqFgUiMSU1JZOLgTCYOzjxSO1RXz9pte1lZGQTEiooPBwQE95VOTU6kd3IiqSkJ9E5KJDUlfB0+svulMHP8QE4fnqNB7B5GYwYicaoxIFZV1rBr/2EOHK7nQG3wOBjx/MDheg6Gz7fsPsi+Q3Xkpfdi1sQCLjm5gLGD0qPdFelAGkAWkRYdrK3nL2uqeOqNCl5cW01dgzNmYBqXnByMZQzM0EytsU5hICLHZef7h/nzii08uaySNzfvxgzOGJ7DJScXMGP8QPrqxj8xSWEgIm22oXoff3xzC0+9UcHmnQdITU6kpCjryED1+HydxRQrFAYi0m7uztJNu3j6zS0s3bSLddv3UtcQ/A5J750UBENBBuPy0xlfkMGw/n11n4duRtcZiEi7mRklRdmUFGUDwSD1um37grOYtuyhrHIPv/77Rg7XNQDQNyWRcfkZnFiYwYmDM5lYmKk9iG5KYSAibdYrKZEJhRlMKMw4Uqutb6C8KgiIsso9rKjcw5xFmzj86jsAZPVJZkJhJicVZnBi+FO3EI0+hYGIdKjkxIQjU2ZQEtzrqra+gbXb9rKiYg/LN+9mecVu7llYTXiEiUEZvTmxMIOCzD6k9U4iPTU5+Nk7ibTejc+Dn2m9k0lJSohiD+OTxgxEJCr2H66jbEsNyzfvZkVFcCV1Vc1B3j9c3+K2mX2SOWNEDlNHD+Dc0bn079erC1ocHzRmICLdSp+UJE4pyuaUcPyhUX2Ds+9gHTUHa6k5WMveg3XhI3hec6CWTTv38+Laav5vxVbM4MTCTKaOHsB5Y3IZn5+hQes2UBiISLeSmGBk9Ekmo8+xp+xuaAjuJrdwbRV/fauKn/5lHXe9sI6cfr04d3QuU8cM4MyROaRr6u9W0WEiEYkLO/Yd4qV11SxcW81La6uoOVhHUoIxMi+NkQP6MSqvHyPz0hiVl8aQ7D49du4lXWcgIj1GXX0Db2zezYtrqyjbUsP67fuo3H3gyPKUpASG5wYBMSovjRED+jFyQD8Ks/rE/eC0xgxEpMdISkz4yHjEvkN1lFftY932vazfvpf1Vfso3RhcQNfIDPIzUhmcncqQ7D7Bo3/fI8+z+iTH7TUSLYaBmT0EXARUufv4sHY78EngMPA28M/uvjtcdjNwFVAPfNXd54f1GcDPgETgAXe/NawPA+YC/YGlwBfc/XAH9lFEhH69kj4y9TcEIbF++17ern6fzTv38274WLi2muq9hz7yHoOz+1CYlUpBZvDIz0ylICuV/Mze5PTtFbOD1y0eJjKzs4F9wJyIMJgO/NXd68zsNgB3v8nMioFHgVOBfOAFYFT4VuuAaUAFsAS4zN1Xm9ljwJPuPtfMfgksd/d7W2q4DhOJSGfbf7iOil0HeHfHByHx7s79VO46QOXuA+w7VPeh9VOSEsjP6B2EQ0YqhVl9GD6gLyMG9GNYTl96JSVGqScfaPNhInd/2cyKmtSej3i5CPhM+HwWMNfdDwHvmFk5QTAAlLv7hrAxc4FZZrYGmAp8PlznYeC7QIthICLS2fqkJDEqHHRuzp4DtWzZfYDKXQfYsufAkZDYsvsAL6+vpmrvoSO3KE0wGJLdh+G5/RgxoB/DB4Q/c/uRkRr9M546Yszgi8Dvw+cFBOHQqCKsAWxuUj+N4NDQbneva2b9jzCzq4GrAYYMGdLuhouItEdGajIZqclHvQHQgcP1bHhvH29Xv0951T7ertpHedU+Xln/HofrG46sl5vWizED0xiXH0zyNy4/naIunuSvXWFgZt8G6oBHOqY5x+bu9wH3QXCYqCs+U0SkrVLDifrG5Wd8qF7f4GzeuZ/yqn2UVwcBsWZrzYduVdo3JZGxg9LDcMigOD+dUXlpnXa2U5vDwMyuJBhY/rh/MPBQCQyOWK0wrHGU+g4g08ySwr2DyPVFROJSYoJRlNOXopy+nE/ekfrhugbWV+2lbEsNq7fUULZlD48vreDh1zYBkJxojByQxqNfmtLiRXnHq01hEJ4Z9B/AOe6+P2LRPOB3ZnYnwQDySGAxYMDI8MyhSuBS4PPu7ma2kGDMYS4wG3i6rZ0REYllKUkJH9mTaGhwNu3cT9mWPZRtqWFD9T7SUzv+qoDWnFr6KHAukGNmFcAtwM1AL2BBeM7tInf/N3cvC88OWk1w+Ogad68P3+daYD7BqaUPuXtZ+BE3AXPN7AfAG8CDHdg/EZGYlpBgDMvpy7Ccvlx0Yn6nfY6uQBYR6UGOdmppfF93LSIiraIwEBERhYGIiCgMREQEhYGIiKAwEBERFAYiIkIMX2dgZtXApohSDvBelJrTWeKtT/HWH4i/PsVbfyD++tTe/gx199ymxZgNg6bMrLS5CyliWbz1Kd76A/HXp3jrD8RfnzqrPzpMJCIiCgMREYmvMLgv2g3oBPHWp3jrD8Rfn+KtPxB/feqU/sTNmIGIiLRdPO0ZiIhIGykMREQkPsLAzGaY2VozKzezb0a7PR3BzDaa2Uoze9PMYu7GDWb2kJlVmdmqiFq2mS0ws/Xhz6xotvF4HaVP3zWzyvB7etPMLoxmG4+HmQ02s4VmttrMyszsurAek9/TMfoTy99RbzNbbGbLwz59L6wPM7PXw995vzezlHZ/VqyPGZhZIrAOmAZUAEuAy9x9dVQb1k5mthEocfeYvFjGzM4G9gFz3H18WPsxsNPdbw1DO8vdb4pmO4/HUfr0XWCfu98Rzba1hZkNAga5+zIzSwOWAhcDVxKD39Mx+vM5Yvc7MqCvu+8zs2TgVeA64OvAk+4+18x+CSx393vb81nxsGdwKlDu7hvc/TDBvZRnRblNPZ67vwzsbFKeBTwcPn+Y4D9qzDhKn2KWu29192Xh873AGqCAGP2ejtGfmOWBfeHL5PDhwFTg8bDeId9RPIRBAbA54nUFMf4PIOTA82a21MyujnZjOkieu28Nn28D8qLZmA50rZmtCA8jxcQhlabMrAg4GXidOPiemvQHYvg7MrNEM3sTqAIWAG8Du929LlylQ37nxUMYxKsz3X0SMBO4JjxEETc8OD4Z28coA/cCw4GJwFbgJ1FtTRuYWT/gCeB6d6+JXBaL31Mz/Ynp78jd6919IlBIcCRkTGd8TjyEQSUwOOJ1YViLae5eGf6sAp4i+EcQ67aHx3Ubj+9WRbk97ebu28P/rA3A/cTY9xQeh34CeMTdnwzLMfs9NdefWP+OGrn7bmAh8DEg08ySwkUd8jsvHsJgCTAyHF1PAS4F5kW5Te1iZn3DATDMrC8wHVh17K1iwjxgdvh8NvB0FNvSIRp/aYYuIYa+p3Bw8kFgjbvfGbEoJr+no/Unxr+jXDPLDJ+nEpwos4YgFD4TrtYh31HMn00EEJ4q9lMgEXjI3X8Y3Ra1j5mdQLA3AJAE/C7W+mRmjwLnEky3ux24Bfgj8BgwhGD68c+5e8wMyB6lT+cSHH5wYCPwrxHH27s1MzsTeAVYCTSE5W8RHGePue/pGP25jNj9jk4kGCBOJPjj/TF3/374O2IukA28AfyTux9q12fFQxiIiEj7xMNhIhERaSeFgYiIKAxERERhICIiKAxERASFgXRTZuZm9pOI198IJ4XriPf+tZl9puU12/05nzWzNWa2sJllo8zsmXBm0GVm9piZ5ZnZuWb25zZ+3vVm1qf9LZeeSGEg3dUh4FNmlhPthkSKuOqzNa4CvuTu5zV5j97A/wH3uvvIcNqRXwC57Wze9cBxhUE466+IwkC6rTqCe71+remCpn/Zm9m+8Oe5ZvaSmT1tZhvM7FYzuzycD36lmQ2PeJvzzazUzNaZ2UXh9olmdruZLQknNfvXiPd9xczmAR+ZGt3MLgvff5WZ3RbWvgOcCTxoZrc32eTzwGvu/qfGgru/6O4fujLWgnn4vxHxepWZFYVXqP+fBXPcrzKzfzSzrwL5wMLGPREzm25mr4V7Hn8I5+xpvFfGbWa2DPismX3VgnsArDCzuS18LxKnjuevHJGudg+wwoL7ILTWScBYgqmmNwAPuPupFtzo5CsEfz0DFBHMUTOc4BfoCOAKYI+7n2JmvYC/mdnz4fqTgPHu/k7kh5lZPnAbMBnYRTDT7MXhVaJTgW+4e9ObE40nmGu/rWYAW9z9E2EbMtx9j5l9HTjP3d8L96j+Ezjf3d83s5sI5sD/fvgeO8I9EsxsCzDM3Q81Tn0gPY/2DKTbCmecnAN89Tg2WxLOa3+IYKrfxl/mKwkCoNFj7t7g7usJQmMMwRxQV4TTBb8O9AdGhusvbhoEoVOAF929OpxS+BGgs2eYXQlMC/+6P8vd9zSzzhSgmCDQ3iSYv2ZoxPLfRzxfATxiZv9EsEcmPZDCQLq7nxIce+8bUasj/LdrZglA5C3/IudnaYh43cCH94SbzsPigAFfcfeJ4WOYuzeGyfvt6UQTZQR7Ei050s9QbwB3X0ewp7IS+EF4SKopAxZE9KXY3a+KWB7Zn08Q7IVNApYc57iIxAmFgXRr4QRpjxEEQqONfPDL9B8I7v50vD5rZgnhOMIJwFpgPvDlcBrkxjN++h7rTYDFwDlmlhMOxl4GvNTCNr8DTjezTzQWzOxsMxvfZL2NBL+gMbNJwLDweT6w391/C9zeuA6wF0gLny8CzggPfzXOhDuqaUPCMB3s7guBm4AMoF8L7Zc4pL8AJBb8BLg24vX9wNNmthx4jrb91f4uwS/ydODf3P2gmT1AcChpmZkZUE0LtxN0960W3Cd4IcFf4//n7secTtjdD4SD1j81s58CtQSHaq4jmBG10RMEh63KCA5brQvrE4Dbzawh3PbLYf0+4Dkz2+Lu55nZlcCj4fgHBGMI6/iwROC3ZpYRtv/ucN586WE0a6mIiOgwkYiIKAxERASFgYiIoDAQEREUBiIigsJARERQGIiICPD/Ad0Jul+2VEN8AAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "print(wss)\n",
    "y_axis = []\n",
    "for _ in wss:\n",
    "    y_axis.append(_[0][0])\n",
    "x_axis = list(range(1,31,1))\n",
    "plt.plot(x_axis, y_axis)\n",
    "plt.xlabel(\"Number of Clusters\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "However, looking at our graph there is no clear elbow point to choose the right number of clustes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-11-27T19:10:37.458076Z",
     "iopub.status.busy": "2021-11-27T19:10:37.457850Z",
     "iopub.status.idle": "2021-11-27T19:10:37.463216Z",
     "shell.execute_reply": "2021-11-27T19:10:37.462120Z",
     "shell.execute_reply.started": "2021-11-27T19:10:37.458045Z"
    }
   },
   "outputs": [],
   "source": [
    "def best_silhouette_score(points, kmax):\n",
    "    sil = []\n",
    "    for k in range(2, kmax+1):\n",
    "        kmeans = KMeans(n_clusters = k).fit(points)\n",
    "        labels = kmeans.labels_\n",
    "        score = silhouette_score(points, labels, metric = 'euclidean')\n",
    "        print(score)\n",
    "        sil.append(score)\n",
    "    return sil"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-11-27T19:10:37.465112Z",
     "iopub.status.busy": "2021-11-27T19:10:37.464863Z",
     "iopub.status.idle": "2021-11-27T19:13:53.031139Z",
     "shell.execute_reply": "2021-11-27T19:13:53.030524Z",
     "shell.execute_reply.started": "2021-11-27T19:10:37.465082Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.5888180743121894\n",
      "0.2304752826883286\n",
      "0.18849012588403366\n",
      "0.20174790279773377\n",
      "0.20725109057613877\n",
      "0.20432755261091265\n",
      "0.19187113583776555\n",
      "0.18252222975284335\n",
      "0.18617177222219883\n",
      "0.18215159046511972\n",
      "0.18678796871997685\n",
      "0.19003701934450365\n",
      "0.17282799290157547\n",
      "0.17525714365752165\n",
      "0.17658376658017552\n",
      "0.17181469078787046\n",
      "0.173550313377515\n",
      "0.1754199728937396\n",
      "0.17508257888210088\n",
      "0.17915222176938547\n",
      "0.17974893770562142\n",
      "0.18349248180250766\n",
      "0.18057324405919467\n",
      "0.18551568254633644\n",
      "0.17240363141274592\n",
      "0.17250963731602864\n",
      "0.18407021223969358\n",
      "0.18068327084532151\n",
      "0.17303101130716814\n"
     ]
    }
   ],
   "source": [
    "sil_scores = best_silhouette_score(vect_np, 30)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's try to use the Silhoutte Method to find the right number of clusters"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-11-27T19:20:49.440084Z",
     "iopub.status.busy": "2021-11-27T19:20:49.439755Z",
     "iopub.status.idle": "2021-11-27T19:20:49.570927Z",
     "shell.execute_reply": "2021-11-27T19:20:49.569720Z",
     "shell.execute_reply.started": "2021-11-27T19:20:49.440048Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0.5888180743121894, 0.2304752826883286, 0.18849012588403366, 0.20174790279773377, 0.20725109057613877, 0.20432755261091265, 0.19187113583776555, 0.18252222975284335, 0.18617177222219883, 0.18215159046511972, 0.18678796871997685, 0.19003701934450365, 0.17282799290157547, 0.17525714365752165, 0.17658376658017552, 0.17181469078787046, 0.173550313377515, 0.1754199728937396, 0.17508257888210088, 0.17915222176938547, 0.17974893770562142, 0.18349248180250766, 0.18057324405919467, 0.18551568254633644, 0.17240363141274592, 0.17250963731602864, 0.18407021223969358, 0.18068327084532151, 0.17303101130716814]\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXQAAAEGCAYAAAB1iW6ZAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjQuMywgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/MnkTPAAAACXBIWXMAAAsTAAALEwEAmpwYAAAgbklEQVR4nO3deZhcdZ3v8fe3qqv3LUtn7ZCEJGSRIGAExQ0UlO2CjrigPupcZ1wRHEev4J3LdZgZr8joxRnR+wDi6FVERr0aJQMuAwqoIQFCIAkJIQvphKQ76fS+Vtf3/nFOdyqd7nSnl1TqnM/reeqpqlOn6/xOTvrTp36/+n2PuTsiIpL/ErlugIiITAwFuohIRCjQRUQiQoEuIhIRCnQRkYgoyNWGp0+f7gsWLMjV5kVE8tKTTz550N1rhnotZ4G+YMEC1q9fn6vNi4jkJTPbPdxro+pyMbNLzWyrmW03sxuHWefdZrbZzDaZ2b1jbayIiIzNiGfoZpYE7gAuAeqAdWa22t03Z62zBLgJeJ27HzazGZPVYBERGdpoztDPA7a7+w537wHuA64etM5fA3e4+2EAd6+f2GaKiMhIRhPoc4E9Wc/rwmXZzgDOMLPHzezPZnbpUG9kZh81s/Vmtr6hoWFsLRYRkSFN1NcWC4AlwIXAtcBdZlY9eCV3v9PdV7n7qpqaIQdpRURkjEYT6HuBeVnPa8Nl2eqA1e7e6+47gW0EAS8iIifJaAJ9HbDEzBaaWSHwXmD1oHV+TnB2jplNJ+iC2TFxzRQRkZGMGOjungauAx4CtgD3u/smM7vFzK4KV3sIOGRmm4GHgc+7+6HJaPC6XY189cHnyWRU9ldEJNuoJha5+xpgzaBlN2c9duCz4W1SPbOniW898iIfv3ARlcWpyd6ciEjeyLtaLv0h3tLZm+OWiIicWvIv0EuCQG9WoIuIHCUPAz3oJWrpTOe4JSIip5a8C/Sq8Ay9pUtn6CIi2fIu0Pv70NXlIiJytPwL9BINioqIDCXvAr2iqAAzBbqIyGB5F+iJhFFRVEBLlwZFRUSy5V2gQ9DtojN0EZGj5WWgV5WkNCgqIjJIXgZ6ZXFKX1sUERkkLwNdZ+giIsfKy0CvLCnQTFERkUHyM9DV5SIicoy8DPSqkhQdPX309mVy3RQRkVNGXga6ZouKiBwrTwM9rLioyUUiIgPyMtCrVBNdROQYeRnoumqRiMix8jLQdYYuInKsvAz0Sl3kQkTkGPkZ6ANdLhoUFRHpl5eBXpxKUJhMqMtFRCRLXga6mQXT/9XlIiIyIC8DHYJ+dJ2hi4gckb+BXqyLXIiIZMvfQC9JaaaoiEiWvA30Kl2GTkTkKHkb6JXFBQp0EZEs+Rvo4aCou+e6KSIip4S8DfSqkhTpjNPZ25frpoiInBLyNtA1W1RE5Gh5G+gq0CUicrS8DfQjF7lQoIuIQD4Humqii4gcJW8DXV0uIiJHy9tA14WiRUSOlr+BXhz0oTfrWy4iIsAoA93MLjWzrWa23cxuHOL1D5tZg5ltCG9/NfFNPVpBMkFZYVKDoiIioYKRVjCzJHAHcAlQB6wzs9XuvnnQqj929+smoY3DqlQ9FxGRAaM5Qz8P2O7uO9y9B7gPuHpymzU6VaqJLiIyYDSBPhfYk/W8Llw22DvNbKOZ/cTM5g31Rmb2UTNbb2brGxoaxtDco1UWp9TlIiISmqhB0V8CC9z9LOA3wPeGWsnd73T3Ve6+qqamZtwbrSwp0KCoiEhoNIG+F8g+464Nlw1w90Pu3h0+vRt41cQ07/jUhy4icsRoAn0dsMTMFppZIfBeYHX2CmY2O+vpVcCWiWvi8NTlIiJyxIjfcnH3tJldBzwEJIF73H2Tmd0CrHf31cD1ZnYVkAYagQ9PYpsHVJWkaO1K05dxkgk7GZsUETlljRjoAO6+BlgzaNnNWY9vAm6a2KaNrH+2aFtXmqrS1MnevIjIKSVvZ4rCkdmi6nYREcnzQFeBLhGRI/I60FWgS0TkiPwO9GKdoYuI9MvrQO8fCFUfuohIngf6wKCoZouKiOR3oJcXFZAwdbmIiECeB7qZBdP/1eUiIpLfgQ7BwKjO0EVEIhDoVSrQJSICRCDQK0sKaOnSoKiISN4Huq5aJCISyPtAryxWl4uICEQh0PUtFxERIAKBXlWSoqs3Q3e6L9dNERHJqbwPdM0WFREJ5H+gq4SuiAgQoUBXP7qIxF3+B3qxaqKLiEAEAl1XLRIRCeR9oFeW9F9XVIOiIhJv+R/o6nIREQEiEOjFqSRFBQkFuojEXt4HOmi2qIgIRCXQiws0KCoisReJQA9qomtQVETiLRKBXqkSuiIi0Qj0KvWhi4hEI9BVE11EJCqBHl6Gzt1z3RQRkZyJRKBXlaToyzjtPaqJLiLxFYlA12xREZGIBLoKdImIRCTQB2qiK9BFJMaiEejFOkMXEYlEoFcNXLVIs0VFJL4iEegDNdF1hi4iMRaJQK9Ql4uISDQCPZkwKooKNP1fRGJtVIFuZpea2VYz225mNx5nvXeamZvZqolr4uioQJeIxN2IgW5mSeAO4DJgBXCtma0YYr0K4AZg7UQ3cjQqVUJXRGJuNGfo5wHb3X2Hu/cA9wFXD7HePwC3Al0T2L5RqyxWl4uIxNtoAn0usCfreV24bICZnQvMc/cHjvdGZvZRM1tvZusbGhpOuLHHE5yhK9BFJL7GPShqZgng68DfjrSuu9/p7qvcfVVNTc14N32UKgW6iMTcaAJ9LzAv63ltuKxfBXAm8IiZ7QJeA6w+2QOjlcUpTSwSkVgbTaCvA5aY2UIzKwTeC6zuf9Hdm919ursvcPcFwJ+Bq9x9/aS0eBhVJSnautOk+zInc7MiIqeMEQPd3dPAdcBDwBbgfnffZGa3mNlVk93A0eqfLdqqs3QRiamC0azk7muANYOW3TzMuheOv1knLrtA15Sywlw0QUQkpyIxUxSyC3RpYFRE4ikygX6kJrq6XEQkniIU6EHvkab/i0hcRSbQ1eUiInEXmUDXVYtEJO4iE+ilhUkKEqbZoiISW5EJdDML6rmoy0VEYioygQ5BxcVmfctFRGIqUoGuAl0iEmeRCnR1uYhInEUr0It1GToRia9oBbouQyciMRaxQC+gpbMXd891U0RETrpIBXpVSYqevgzdadVEF5H4iVSg988W1TddRCSOohXoJZr+LyLxFalAV4EuEYmzSAV6ZbFK6IpIfEUq0Kt0kQsRibFIBXqlulxEJMaiFej9NdE7FOgiEj+RCvTCggQlqaTO0EUkliIV6BDMFtWgqIjEUfQCvVj1XEQkniIX6FUqoSsiMRW5QK8sUQldEYmnyAW6ztBFJK4iF+iVxQXqQxeRWIpeoIdn6JmMaqKLSLxELtCrSlK4Q1uPztJFJF4iF+iaLSoicRW9QFc9FxGJqQgGelBCVwOjIhI30Qv0Yl21SETiKXKBrqsWiUhcRS7QB/rQdYYuIjETuUCvKCrATIEuIvETuUBPJIyKogJaujQoKiLxMqpAN7NLzWyrmW03sxuHeP3jZvasmW0ws8fMbMXEN3X0VKBLROJoxEA3syRwB3AZsAK4dojAvtfdV7r72cBXga9PdENPRFVJSl0uIhI7ozlDPw/Y7u473L0HuA+4OnsFd2/JeloG5LSQSmWxKi6KSPyMJtDnAnuynteFy45iZp8ysxcJztCvH+qNzOyjZrbezNY3NDSMpb2josvQiUgcTdigqLvf4e6LgC8AfzfMOne6+yp3X1VTUzNRmz5G0OWiQVERiZfRBPpeYF7W89pw2XDuA94+jjaNW2WxBkVFJH5GE+jrgCVmttDMCoH3AquzVzCzJVlPrwBemLgmnrjKkhSdvX30pDO5bIaIyElVMNIK7p42s+uAh4AkcI+7bzKzW4D17r4auM7MLgZ6gcPAhyaz0SPpn/7f2tXLtPKiXDZFROSkGTHQAdx9DbBm0LKbsx7fMMHtGpf+iovNnQp0EYmPyM0UhewCXRoYFZH4iGSgq4SuiMRRNANdFRdFJIYiGeiqiS4icRTJQFeXi4jEUSQDvTiVoDCZ0GxREYmVSAa6mVFZUqAuFxGJlUgGOmj6v4jET3QDXTXRRSRmFOgiIhER3UAv1nVFRSReIhvougydiMRNZAO9/0LR7jm9Gp6IyEkT2UCvKkmRzjidvX25boqIyEkR2UDXbFERiZvoBnpYE12zRUUkLiIb6CrQJSJxE9lAH+hy6VCgi0g8RDbQdYYuInET2UDvv8iFBkVFJC4iG+gVxRoUFZF4iWygp5IJygqT6nIRkdiIbKDDkdmiIiJxEO1AL1Y9FxGJj0gHelVJSl0uIhIbkQ70ypICmjUoKiIxEfFAT9Hc0ZPrZoiInBSRDvTlsyrZ19zFul2NuW6KiMiki3Sgf+A185lZWcSX12xRXXQRibxIB3pJYZK/ufgMnn6piQef25/r5oiITKpIBzrANa+qZcmMcm598Hl6+zK5bo6IyKSJfKAXJBPceNkydh3q4EdPvJTr5oiITJrIBzrAm5fN4PyFU/nGb1+grVtfYxSRaIpFoJsZN12+nEPtPdz5+xdz3RwRkUkRi0AHOHteNVeeNZu7Ht3JgZauXDdHRGTCxSbQAT7/tqWkMxlu/+22XDdFRGTCxSrQ508r4/3nz+fH6/bwwoHWXDdHRGRCFeS6ASfbp9+8mJ8+WcetD27l7g+tmrTtvNzcyRM7G3lq92EKCxLMm1pK7ZQS5k0ppXZKKSWFyUnbtojE06gC3cwuBb4BJIG73f0rg17/LPBXQBpoAP6ru++e4LZOiGnlRXz8wkXc9tBW1u44xPmnTxv3e7o7Ow62s25nI0/sauSJnY3UHe4EoKwwSTrjdKeP/g789PJCaqeEIT+1lHlTSjltainnLZxKYUGsPjiJyASxkabEm1kS2AZcAtQB64Br3X1z1joXAWvdvcPMPgFc6O7vOd77rlq1ytevXz/e9o9JZ08fF/3zI8ysKubnn7wAMzuhn+/LOFtebmFdGN7rdjVysC0oAjatrJDzFk7l1Qumct7CqSyfXUnCoKGtmz2NndQd7qDucCd7GsP7wx3sa+qkty84DqdNLeVzb1vKlStnk0icWLtEJPrM7El3H7J7YTRn6OcB2919R/hm9wFXAwOB7u4PZ63/Z+ADY2/u5CspTPLZt57Bf/vJRtY8u58rzpo9qp9zd9Y8u58vr9nC3qbgDLx2SglvXFLDqxcGAX769LIh/0DMqChmRkUxr5o/5ZjX+jLOgZYuNtY1c/tvt3H9j57mrj/s4KbLlnHB4unj21kRiY3RBPpcYE/W8zrg/OOs/xHgP8bTqJPhnefW8p1Hd/LVh57nkhUzR+zm2LSvmb//5Wae2NnIslkV/O/3vJLzF05jTnXJuNuSTBhzqkuYU13CJStm8vOn9/L132zjfXev5U1n1HDjZctYPrty3NsRkWib0M5aM/sAsAq4bZjXP2pm681sfUNDw0Ru+oQlE8aNly1j96EO7l07fHf/obZubvrZs1z5r4+xvb6NL79jJQ9c/wbecU7thIT5UO1656tq+d3fvokvXr6MDXuauPxfHuWz928Y+FQgIjKU0fShvxb4kru/LXx+E4C7/69B610M/CvwJnevH2nDuexD7+fuvO+utWw90MrvP38hFcWpgdd60hm+/6ddfON3L9DZ08cHX7uAGy5eQlVJ6jjvOPGaO3r51iPb+e4fdwHwlxcs4JMXLqaq9OS2Q0RODcfrQx9NoBcQDIq+BdhLMCj6PnfflLXOOcBPgEvd/YXRNOpUCHSAjXVNXPXNx7nuosV87m1LAXh4az3/8KvN7Gho541n1HDzlctZPKMip+3c29TJ13+9jZ89XUdlcYpPXriI979mPuVFsfvmqUisjSvQwze4HLid4GuL97j7P5nZLcB6d19tZr8FVgIvhz/ykrtfdbz3PFUCHeD6Hz3Nrzfv554Pv5q7/rCDh7c2sHB6Gf/jyuVctHTGCX8LZjJtebmFWx98nke2NlBWmOTqc+by/vNP4xVzqnLdtDHr7cuwp7GD3Yc62HO4g1Xzp7JijsYMRIYy7kCfDKdSoO9p7ODNX3uE3j6noqiA69+yhA9dsOCU/j74hj1N/ODPu/nlM/voTmc457Rq3n/+fK48azbFqVNv0lJ3uo89jZ3sPtTOrkMd7DrYzq5D7ew+1MHepk76Mkf+H6aSxk2XLecvX7fglPpjKnIqUKCPwv/90y5ebGjnUxctpqaiKNfNGbXmjl5++lQdP1y7mxcb2qkqSXHNq2p53/mnsaimfMSfb+3qZXt9Gy82tLO9vo3mzl4Wzyhn2awKls6qYHr5if9btHWn2bS3mWf3NrOxLrjffaidrMymoqiABdPLgtu0UuZPC+6nlxfxjw9s4bdbDvDWFTO57ZpXarwgZvoyzgv1rbR2pVk1f4r+qA+iQI8Bd+fPOxr54drdPLRpP719zgWLpvH+8+dzyYqZHO7o4cX6NrY3tLG9vi0M8TYOtHQPvEcqaZQVFdDU0TuwbHp5IUtnVbB0ZuVAyJ8xs2KgdEFHT5rN+1oGgntjXRM7DrbT/99qTlUxZ86tYtmsiqwAL2NKaWrYX1R35zuP7eQr//E8s6qKueN95/LKedWT9m8nY9PWnebF+jZeauxgWlkw83l2dTGp5Il9sq1v6eLpPU08/VITG/Yc5tm6Ztp7+gB49YIp3HzlK1hZe3K7FA+39/C9P+2iqaOXhdPLBm5zqktI5njCnwI9Zhpau7l//R7uXfsSe5s6SSbsqC6N8qICFs0oZ3FNOYtmlLG4ppzFM8o5bWopBckEDa3dbN3fyvP7W9i6v5VtB1rZdqCNzt7gl8wM5k8tpbAgwfb6toEz75mVRaycW83KuVWcVVvFmXOrxvVp56mXDvPpe5+mvrWLL16+nA9foC6YXGju6OWF+la217fxQnjbfqCVfc3HlqFOGMyuKhkoaXGkflHwvLIkxeZ9LWzYc5gNe5rY8FLTwPukksby2ZWcPa+as+dV097Tx+2/2UZjRw/XnFvL59+2lBmVxZO6r00dPdz16A7+7fFddPT2UZpKDvxxASgsSDB/amkQ8DVlnD69jIXTy1k4vYzp5YUn5f+nAj2m+jLOH7Y18McXD1I7pZTFM8pZVFPOzMqiMZU72NPYwfP7W9m6v5WtB1ro7s1wZhjeK+dWTcovW1NHD5/792f47ZZ63vaKmXz1mlee9K+ORkH/H+nmzl56+vroSWfo6fPgPp2hty8TLgvuu9N97D7UwQv1bTS0HvkUV5xKsHhGOUtmVLB4RnAiMH9aKY3tPdSFpS32HA7vGzs50NrFcBFTO6WEc06bMhDgr5hTecz4T0tXL3f853bueXwnqWSCT120mI+8fuGEjxM1d/Ry92M7+O7ju2jvSXP5ytnc8JYlLJlRTkNrNzsOtrPzYDu7DrYPPN59qH2gZAfA6TVlXLlyNpefNZulMysmLdwV6JLXsrtgZlcX881rJ6cLxt3p7A3Crqpk+C6hU1lnTx8v1Lfy/MutPJ/1KetQe8+IP2sGhckEhQUJigqS1E4pYcmMcpbMPBLgc6tLTqjGUHe6j31NXQO1iw539LB0ZgVnn1Z9QuMzuw628+U1W/j15gPUTinhpsuWc/nKWeM+Rs2dvXznsZ1897GdtHanuXzlLG54yxksnTXy15TTfRn2NXWx42DQhfm7LfWs3XmIjMOimjKumKRwV6BLJGR3wfz3y5fzoeN0wXT29LG/pYv9zV3sb+mkvqWb5s5eWrvStHb10tadpqUrPfC8tStNW3d6oGtqRkURZ9VW88raKs6aF9xXlxZO6P6k+zI0dvRwuL2XQ+3dNLb30Njew6G2Hrp6+0gmjIKEkQjvk4lEeH/0rb6leyC4dx46Mn5RnEqwdGYw7rFsVjAGMq28iMKCBKmkBcGdTJIqMAqTCQpOsO87F/64/SC3/Gozz+9v5bwFU7n5v6zgzLkn3r/e0tXLPY/t5DuP7aS1K82lr5jFDRcvGXeJjYbWbh7ctJ81G18+JtyvOGsOZ8wsH3e4K9AlMrK7YC59xSzecMZ0DjR3BeHd0s3+5k72N3fR0nXsxcCTCaOiuCC4FaWOPC4++nHCYMvLrTxT18SOhvaBn58/rfRIyNdWc+bcSkoLg4ldfRmnubOXwx09HG7v4XBH8Lipo4fG9t7w/sjtUHsPzZ29x7SxX1FBgow76YwP22XRr39MY9msSpbOqmD57AqWzqrktKmlOR/Amwx9GefH6/bwtV9vHehf/9ibFlGcSuBOcMPDe8h4/7+hk3F48Ln93P3oDlq60rx1xUxuuHjJpMzjGDbcz5rDX5wzlwXTy8b0vgp0iZTsLph0xjGDmvIiZlUVM6uymFlVxcysPPrxzMoiyosKTvjsqKWrl+fqmtlQ18TGPcG3ePoH8RIGc6pLaOtO09zZO2zwFiYTVJemmFpWeMxtWlkhU8uKgsflwbLqktRRZ8uZjNPnTl8mCPi+vuB5OpMh3edUl6YG/rDESUtXL9/8z+189/GdR/Vlj8bFy2fymYuXjOnsfiz6w/2Bjft4YmcjX37HSt573mljei8FukRSfWsXfRmnprzopHYX1Ld2DYT77sYOqkpSVJcWMiUM7erSQqaWFlJdmmJKWSFlhcm87I/PF7sPtbN2RyMYGGBm4T0kzOj/p+9fvqimPKczketbuygtLBhz2Q4FuohIRBwv0E/9URARERkVBbqISEQo0EVEIkKBLiISEQp0EZGIUKCLiESEAl1EJCIU6CIiEZGziUVm1gDszsnGJ8Z04GCuGzEJorpfEN19037ln/Hs23x3rxnqhZwFer4zs/XDzdbKZ1HdL4juvmm/8s9k7Zu6XEREIkKBLiISEQr0sbsz1w2YJFHdL4juvmm/8s+k7Jv60EVEIkJn6CIiEaFAFxGJCAX6GJjZLjN71sw2mFneXqXDzO4xs3ozey5r2VQz+42ZvRDeT8llG8dimP36kpntDY/ZBjO7PJdtHAszm2dmD5vZZjPbZGY3hMujcMyG27e8Pm5mVmxmT5jZM+F+/X24fKGZrTWz7Wb2YzObkCuQqw99DMxsF7DK3fN60oOZvRFoA77v7meGy74KNLr7V8zsRmCKu38hl+08UcPs15eANnf/51y2bTzMbDYw292fMrMK4Eng7cCHyf9jNty+vZs8Pm4WXHuwzN3bzCwFPAbcAHwW+Jm732dm/wd4xt2/Pd7t6Qw9xtz9D0DjoMVXA98LH3+P4JcqrwyzX3nP3V9296fCx63AFmAu0Thmw+1bXvNAW/g0Fd4ceDPwk3D5hB0zBfrYOPBrM3vSzD6a68ZMsJnu/nL4eD8wM5eNmWDXmdnGsEsm77olspnZAuAcYC0RO2aD9g3y/LiZWdLMNgD1wG+AF4Emd0+Hq9QxQX+8FOhj83p3Pxe4DPhU+BE/cjzoj4tKn9y3gUXA2cDLwNdy2ppxMLNy4KfAZ9y9Jfu1fD9mQ+xb3h83d+9z97OBWuA8YNlkbUuBPgbuvje8rwf+H8FBiooDYX9mf79mfY7bMyHc/UD4i5UB7iJPj1nYD/tT4Ifu/rNwcSSO2VD7FpXjBuDuTcDDwGuBajMrCF+qBfZOxDYU6CfIzMrCQRvMrAx4K/Dc8X8qr6wGPhQ+/hDwixy2ZcL0B17oHeThMQsH2L4DbHH3r2e9lPfHbLh9y/fjZmY1ZlYdPi4BLiEYH3gYuCZcbcKOmb7lcoLM7HSCs3KAAuBed/+nHDZpzMzsR8CFBKU8DwD/E/g5cD9wGkF543e7e14NMA6zXxcSfGx3YBfwsax+57xgZq8HHgWeBTLh4i8S9DXn+zEbbt+uJY+Pm5mdRTDomSQ4gb7f3W8Jc+Q+YCrwNPABd+8e9/YU6CIi0aAuFxGRiFCgi4hEhAJdRCQiFOgiIhGhQBcRiQgFukwKM3Mz+1rW88+FBbIm4r3/zcyuGXnNcW/nXWa2xcweHuK1M8xsTVjh8Ckzu9/MZprZhWb2qzFu7zNmVjr+lktcKdBlsnQDf2Fm03PdkGxZs/NG4yPAX7v7RYPeoxh4APi2uy8Jy0B8C6gZZ/M+A5xQoJtZcpzblAhRoMtkSRNcN/FvBr8w+AzbzNrC+wvN7Pdm9gsz22FmXzGz94f1pJ81s0VZb3Oxma03s21mdmX480kzu83M1oXFnD6W9b6PmtlqYPMQ7bk2fP/nzOzWcNnNwOuB75jZbYN+5H3An9z9l/0L3P0Rdz9qFmNYy/tzWc+fM7MF4WzjB8Ia2c+Z2XvM7HpgDvBw/ycCM3urmf0p/ATw72Gdk/56/Lea2VPAu8zsegvqiG80s/tGOC4SYSdytiJyou4ANlpQY320XgksJyh/uwO4293Ps+CCB58mOIsFWEBQ12MRQQguBj4INLv7q82sCHjczH4drn8ucKa778zemJnNAW4FXgUcJqii+fZwNt+bgc+5++CLmJxJUK97rC4F9rn7FWEbqty92cw+C1zk7gfDTzZ/B1zs7u1m9gWCGtq3hO9xKPxkgJntAxa6e3f/NHOJJ52hy6QJq+V9H7j+BH5sXVgbu5ugzGh/ID9LEOL97nf3jLu/QBD8ywjq6nwwLFW6FpgGLAnXf2JwmIdeDTzi7g1hOdMfApNdPfNZ4JLwLPsN7t48xDqvAVYQ/FHaQFDvY37W6z/OerwR+KGZfYDgk5HElAJdJtvtBH3RZVnL0oT/98wsAWRffiu7nkUm63mGoz9RDq5Z4YABn3b3s8PbQnfv/4PQPp6dGGQTwRn9SAb2M1QM4O7bCD4xPAv8Y9i9M5gBv8nalxXu/pGs17P35wqCT0PnAutOcJxAIkSBLpMqLBJ1P0Go99vFkUC8iuAqLifqXWaWCPvVTwe2Ag8BnwjLsPZ/E6XseG8CPAG8ycymhwOM1wK/H+Fn7gUuMLMr+heY2RvN7MxB6+0iCFnM7FxgYfh4DtDh7j8AbutfB2gFKsLHfwZeF3Yl9Vf5PGNwQ8I/iPPc/WHgC0AVUD5C+yWi9JdcToavAddlPb8L+IWZPQM8yNjOnl8iCONK4OPu3mVmdxN0yzxlZgY0MMKlvdz9ZQuuw/kwwVnxA+5+3FKm7t4ZDsTebma3A70E3R43EFR47PdTgi6gTQRdQNvC5SuB28wsE/7sJ8LldwIPmtk+d7/IzD4M/CgcD4CgT30bR0sCPzCzqrD9/xLW3ZYYUrVFEZGIUJeLiEhEKNBFRCJCgS4iEhEKdBGRiFCgi4hEhAJdRCQiFOgiIhHx/wEjppPZx+2YfQAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "print(sil_scores)\n",
    "x_axis = list(range(2,31,1))\n",
    "plt.plot(x_axis, sil_scores)\n",
    "plt.xlabel(\"Number of Clusters\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "But here also there is no clear peak with which we can choose the Number of Clusters. Let's explore why clustering is not a good option here, with k=30."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-11-27T19:13:53.207876Z",
     "iopub.status.busy": "2021-11-27T19:13:53.207362Z",
     "iopub.status.idle": "2021-11-27T19:13:59.704570Z",
     "shell.execute_reply": "2021-11-27T19:13:59.703886Z",
     "shell.execute_reply.started": "2021-11-27T19:13:53.207838Z"
    }
   },
   "outputs": [],
   "source": [
    "kmeans = KMeans(n_clusters = 30).fit(vect_np)\n",
    "training_labels = kmeans.labels_\n",
    "testing_labels = kmeans.predict(vect_test_np)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-11-27T19:13:59.705951Z",
     "iopub.status.busy": "2021-11-27T19:13:59.705572Z",
     "iopub.status.idle": "2021-11-27T19:13:59.830262Z",
     "shell.execute_reply": "2021-11-27T19:13:59.829637Z",
     "shell.execute_reply.started": "2021-11-27T19:13:59.705919Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(794, 33)\n",
      "Round 0\n",
      "(544, 33)\n",
      "Round 1\n",
      "(255, 33)\n",
      "Round 2\n",
      "(482, 33)\n",
      "Round 3\n",
      "(916, 33)\n",
      "Round 4\n",
      "(500, 33)\n",
      "Round 5\n",
      "(653, 33)\n",
      "Round 6\n",
      "(747, 33)\n",
      "Round 7\n",
      "(59, 33)\n",
      "Round 8\n",
      "(318, 33)\n",
      "Round 9\n",
      "(325, 33)\n",
      "Round 10\n",
      "(242, 33)\n",
      "Round 11\n",
      "(197, 33)\n",
      "Round 12\n",
      "(266, 33)\n",
      "Round 13\n",
      "(205, 33)\n",
      "Round 14\n",
      "(223, 33)\n",
      "Round 15\n",
      "(381, 33)\n",
      "Round 16\n",
      "(190, 33)\n",
      "Round 17\n",
      "(891, 33)\n",
      "Round 18\n",
      "(98, 33)\n",
      "Round 19\n",
      "(332, 33)\n",
      "Round 20\n",
      "(236, 33)\n",
      "Round 21\n",
      "(438, 33)\n",
      "Round 22\n",
      "(352, 33)\n",
      "Round 23\n",
      "(824, 33)\n",
      "Round 24\n",
      "(336, 33)\n",
      "Round 25\n",
      "(1388, 33)\n",
      "Round 26\n",
      "(94, 33)\n",
      "Round 27\n",
      "(207, 33)\n",
      "Round 28\n",
      "(907, 33)\n",
      "Round 29\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>Tokens</th>\n",
       "      <th>Processed Text</th>\n",
       "      <th>Folder</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>[hour]</td>\n",
       "      <td>hour</td>\n",
       "      <td>mann-k</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>37</th>\n",
       "      <td>[yes]</td>\n",
       "      <td>yes</td>\n",
       "      <td>mann-k</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>51</th>\n",
       "      <td>[know, either, kay]</td>\n",
       "      <td>know either kay</td>\n",
       "      <td>mann-k</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>54</th>\n",
       "      <td>[speechless]</td>\n",
       "      <td>speechless</td>\n",
       "      <td>mann-k</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>60</th>\n",
       "      <td>[anything, get, set, booked, 200, otherwise, a...</td>\n",
       "      <td>anything get set booked 200 otherwise availabl...</td>\n",
       "      <td>mann-k</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13312</th>\n",
       "      <td>[ces, synthetic, storage, deal, equitable, gas...</td>\n",
       "      <td>ces synthetic storage deal equitable gas compa...</td>\n",
       "      <td>germany-c</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13314</th>\n",
       "      <td>[sure, one, gas, purchase, make, right, 9, 00,...</td>\n",
       "      <td>sure one gas purchase make right 9 00 every mo...</td>\n",
       "      <td>germany-c</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13324</th>\n",
       "      <td>[please, add, tricia, spence, email, working, ...</td>\n",
       "      <td>please add tricia spence email working cgas vo...</td>\n",
       "      <td>germany-c</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13375</th>\n",
       "      <td>[read, call, make, brownies, bring]</td>\n",
       "      <td>read call make brownies bring</td>\n",
       "      <td>germany-c</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13399</th>\n",
       "      <td>[changed, term, deal, ticket]</td>\n",
       "      <td>changed term deal ticket</td>\n",
       "      <td>germany-c</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>907 rows × 3 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                  Tokens  \\\n",
       "7                                                 [hour]   \n",
       "37                                                 [yes]   \n",
       "51                                   [know, either, kay]   \n",
       "54                                          [speechless]   \n",
       "60     [anything, get, set, booked, 200, otherwise, a...   \n",
       "...                                                  ...   \n",
       "13312  [ces, synthetic, storage, deal, equitable, gas...   \n",
       "13314  [sure, one, gas, purchase, make, right, 9, 00,...   \n",
       "13324  [please, add, tricia, spence, email, working, ...   \n",
       "13375                [read, call, make, brownies, bring]   \n",
       "13399                      [changed, term, deal, ticket]   \n",
       "\n",
       "                                          Processed Text     Folder  \n",
       "7                                                   hour     mann-k  \n",
       "37                                                   yes     mann-k  \n",
       "51                                       know either kay     mann-k  \n",
       "54                                            speechless     mann-k  \n",
       "60     anything get set booked 200 otherwise availabl...     mann-k  \n",
       "...                                                  ...        ...  \n",
       "13312  ces synthetic storage deal equitable gas compa...  germany-c  \n",
       "13314  sure one gas purchase make right 9 00 every mo...  germany-c  \n",
       "13324  please add tricia spence email working cgas vo...  germany-c  \n",
       "13375                      read call make brownies bring  germany-c  \n",
       "13399                           changed term deal ticket  germany-c  \n",
       "\n",
       "[907 rows x 3 columns]"
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "for i in range(0,30,1):\n",
    "    cluster_df = df.iloc[np.where(training_labels==i)]\n",
    "    print(cluster_df.shape)\n",
    "    print(\"Round {}\".format(i))\n",
    "cluster_df[[\"Tokens\", \"Processed Text\", \"Folder\"]]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "It looks like the data is getting pretty nicely distributed amongst the 30 clusters. Here clustering is based on the vectors of the emails - which is the average of all the word vectors contained in each email. Let's look at the emails found within the last cluster. As you can see from the Processed Text none of the emails really have anything in common, they show no proper relationship or meaningful insight. \n",
    "We believe that by taking the average of all the word vectors and using that to represent the overall email lost a lot of information, and thus clustering using the resulting vector did not result in meaningful clusters."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.8.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}