1. Environment settings

Run pip install -r requirements.txt

2. Place challenge data

1. Place challenge data at ./data/physionet2019/raw/.
2. Unzip challenge data.

3. Data preparation

1. cd data_prep/

2. Run python 01_makedataset.py

3. Run python 02_convert_data.py <scaling> <imputation> <preprocess> <seed>

   Choose preprocess type, imputation type and scaler type.

   • : Scaling function to call from custom_scalers.py

     - 0: no scaling
     - 1: standard scaler (default)
     - 2: custom
     

   • : Imputation function to call from imputation_functions.py

     - 0: mean imputation (default)
     - 1: forward imputation
     

   • : Preprocess function to call from manual_preprocessor.py

     - 0: No preprocess
     - 4: Dafault
     - For details, please check `manual_preprocessor.py`
     

   • : Number of random seed to split data (Default: 1)

4. model training

1. cd prediction/
2. python train.py (By default hyper parameter settings, model achieves following normalized utility score.)

   train,valid,test
   0.431469,0.415992,0.420764