MAIF Challenge

My practice codes for MAIF challenge, which is open-competition of machine learning and data science. This practice was managed by Machine Learning course in 2016 HEIG-VD & SNU Summer University, Switzerland.

Short Description

Infer the annual insuarance fee from other given variables(known or unknown).

• data/ech_apprentisage.csv Training Data
• data/ech_test.csv Test Data

Shotgun & Ensemble Method

Using all given variables, train several models and merge the results using simple regressor. String variables are converted to binary strings.

PATH_TO_ANACONDA/bin/python shotgun_and_ensemble.py first[second]

The best result : 12.476 %

Many Decision Trees Method

The experimental method which makes many decision trees, trained by randomly-selected samples from original training data, and merge their decision.

PATH_TO_ANACONDA/bin/python many_decision_trees.py

The best result : 16.557 %