techvick / kgl_mds

we gonna win this hackathon

KGL_MDS

1. mds_s1 ordinal encoding - converting feature variables to their corresponding integer value Decision tree regressor with depth 2

2. mds_s2 ordinal encoding - converting feature variables to their corresponding integer value RandomForestRegressor and GradientBoostingRegressor, with a internal commented loop for tree depth and other parameter optimization using grid search

3. mds_s3 ordinal encoding - converting feature variables to their corresponding integer value XGBRegressor

4. mds_s4 One hot encoding - total variable in total column and each with one and rest zero RandomForestRegressor,with a internal commented loop for tree depth and other parameter optimization

5. mds_s5 empty for now

6. mds_s6 One hot encoding - total variable in total column and each with one and rest zero RandomForestRegressor with decomposition PCA for feature reduction np.setdiff1d is used to match train and test data sets after feature reduction, clumns should be same. Train test can have different features, so this is necessary.

7. mds_s7 ordinal encoding - converting feature variables to their corresponding integer value RandomForestRegressor with decomposition PCA for feature reduction

8. mds_s8 Binary encoding : feature to integer, integer to binary, binary to columns AdaBoostRegressor, GradientBoostingRegressor, with a internal commented loop for tree depth and other parameter optimization