Quick Start

You can setup the environment for running this repo quickly with conda.

1. Create the environment.

conda env create -f environment.yml

2. Activate the environment.

conda activate cs5228_project

How to Run

1. Run EDA/EDA.ipynb
2. Run DataPreprocess.ipynb
3. (Optional) Run FineTuning.ipynb: run the notebook if you want to check out the fine tuning steps.
4. Run DataMining.ipynb

File Structure

• data/*

  Original training and test data, preprocessed training and test data and prediction results from different models.

  The final train file after preprocessed is the train_preprocessed_impute.csv, optimized from the train_preprocessed.csv by applying IterativeImputer of sklearn.impute.

  The final text file after preprocessed is the test_preprocessed.csv.

  The different /*result.csv are prediction results from different models. The final best kaggle scores are from the xgb_result_best.csv.

• EDA/EDA.ipynb

  Jupiter notebook containing the steps for EDA.

• images/*

  Images used by the jupiter notebooks in this repository or the final report.

• util/DataMining.py

  Util functions used by the DataMining.ipynb jupiter notebook.

• util/DataPreprocess.py

  Util functions used by the DataPreprocess.ipynb jupiter notebook.

• DataMining.ipynb

  Jupiter notebook containing the data mining and prediction steps for XGB and XGB by make models.

• DataPreprocess.ipynb

  Jupiter notebook containing the preprocess steps for the training and the Kaggle test dataset.

• FineTuning.ipynb

  Jupiter notebook containing the hyperparameter fine_tuning steps of partial models, and some exploratory work for feature selections.