arnabbiswas1 / k_tab_aug_muticlass_rmse_logloss_weightedf1_stratified_tsfresh_cesium Goto Github PK

Repository for Kaggle TPS August 2021

1. Clone the source code from github under <PROJECT_HOME> directory.

git clone https://github.com/arnabbiswas1/kaggle_tab_jun.git

This will create the following directory structure:

<PROJECT_HOME>/kaggle_tab_jun

2. Create conda env:

conda env create --file environment.yml

3. Go to the raw data directory at <PROJECT_HOME>/kaggle_tab_jun/data/raw. Download dataset from Kaggle:

kaggle competitions download -c tabular-playground-series-jun-2021

4. Unzip the data:

unzip tabular-playground-series-jun-2021.zip

5. Set the value of variable HOME_DIR at <PROJECT_HOME>/kaggle_tab_jun/src/config/constants.py with the absolute path of <PROJECT_HOME>/kaggle_tab_jun

6. To process raw data into parquet format, go to <PROJECT_HOME>/kaggle_tab_jun/src. Execute the following:

python -m scripts.process_raw_data

7. To trigger feature engineering, go to <PROJECT_HOME>/kaggle_tab_jun/src. Execute the following:

python -m scripts.create_fetaures

Following is needed for visualizing plots for optuna using plotly (i.e. plotly dependency):

jupyter labextension install [email protected]