WSDM2022-Retention-Score-Prediction

WSDM2022留存预测挑战赛 第1名解决方案

赛题链接：http://challenge.ai.iqiyi.com/detail?raceId=61600f6cef1b65639cd5eaa6

解决方案blog：https://zhuanlan.zhihu.com/p/462736790?

1. 环境配置

• python--3.8
• scipy==1.7.1
• pandas==1.3.2
• numpy==1.19.5
• lightgbm==2.0.0
• catboost==0.25.1
• xgboost==1.2.1
• torch==1.9.0+cu111
• deepctr-torch==0.2.7
• tensorflow-decison-forests==0.2.2

2. 目录结构

./
├── README.md
├── get_features.ipynb
├── get_sequence.ipynb
├── cnn_binary_0.ipynb
├── cnn_binary_7.ipynb
├── tfdf_online.ipynb
├── inference.ipynb
├── function.py

4. 运行流程

• 安装环境
• 创建data、features和res文件夹，将a榜和b榜数据集解压后放到data目录下
• 依次运行get_features.ipynb，get_sequence.ipynb，cnn_binary_0.ipynb，cnn_binary_67.ipynb， tfdf_online.ipynb， inference.ipynb，最终结果生成在在res文件夹中