最终B榜成绩0.7849(auc)
本地克隆
git clone https://github.com/cbwces/CMB-Fintech2020.git
解压数据
cd ./CMB-Fintech2020
tar -xjvf ./train.tar.bz2 && tar -xjvf ./test.tar.bz2
运行
python3 ./fintech.py
-
fintech.py: 主要执行文件,数据预测
-
训练数据集tag.csv、 训练数据集beh.csv、 训练数据集trd.csv: 分别对应用户特征、app记录、交易记录的相关训练数据
-
评分数据集tag.csv、 评分数据集beh.csv、评分数据集trd.csv: 相应训练集
-
null_importances_distribution_rf.csv、 actual_importances_distribution_rf.csv: null importance特征筛选评估表,提供特征筛选依据方案来源
缺少相应第三方库,可通过pip install -r ./requirements.txt
进行安装
默认关闭GPU加速,如需要进行GPU加速,将fintech.py脚本内模型相应参数激活
各csv文件默认与执行脚本路径位置相同,可以通过变量进行传参
传参形式: python3 ./fintech.py --训练集路径 --测试集路径 --特征路径