American-Express---Default-Prediction

Competitions from the kaggle platform: https://www.kaggle.com/competitions/amex-default-prediction

Build

$ git clone https://github.com/zeio99/American-Express---Default-Prediction.git
$ cd code

1. Steps to run

• 1 Download the raw dataset and save it to the input/amex-default-prediction directory. Download the processed dataset to the input/amex-data-integer-dtypes-parquet-format directory (https://www.kaggle.com/datasets/raddar/amex-data-integer-dtypes-parquet-format)
• 2 Run code/fe_process.py, the main purpose of this step is to generate the feature file, which takes some time to run
• 3 Run code/lgb.py, this code trains lgb
• 4 Run code/xgb.py, this code trains xgb (using GPU)
• 5 Run code/lgb_2.ipynb, which trains the second lgb
• 6 Run code/infer.ipynb to get the fusion results

2. Prerequisites

• pandas
• numpy
• lightgbm
• pyarrow
• pickle
• tdqm