First place solution for How Much Did it Rain?

For a description of my solution see this blog post on kaggle, or this forum post

Running

This should run with python 3 and the newest version of xgboost.The following folders should be in the top level directory, input, code, processed, models, output. Look at documentation.pdf for required directory structure and other information on running the code. Put train_2013.csv and test_2014.csv into the input folder. Run ./main.sh in the code folder.

For faster run time

Simple Features and methods In train_subset4.py and train_subset5.py change the line use_xtra_features = True to False.

In main.sh comment out the lines

./make_xtra_features_train.sh and ./make_xtra_features_test.sh.

In make_predictions.py change the line with

fn.make_prediction4(......use_xtra_features=True) to fn.make_prediction4(.....use_xtra_features=False),

repeat for

fn.make_prediction5(....)

Now run main.sh and it should shave off ~5 hours of run time. Final CRPS of about 0.00750.