Yet Another KDD Cup 2015 Solution
Based on
Final result
Public board
This leaderboard calculates scores by about 50% of the test data.
AUC score: 0.8874428893001793
Rank: #89
Private board
AUC score: 0.8867360110438458
Rank: #86
TODO
加特征
加数据:最多可以到depth=4
feature selection
观察预测错了的instance
尝试深度学习model,
kNN
GBDT: min_samples_leaf, max_depth ...
RF: min_samples_leaf, max_depth ...
AdaBoost(with LR/SVC
2022-05-05 00:44:22
72KB
Python
1