Awesome

Santander Product Recommendation - 8th place

Caution

• make_data() step in main.py needs 30GB of memory but it can be optimized.

This code produces 3 submissions

• xgboost - 0.03061 public LB
• lightgbm - 0.03059 public LB
• xgboost+lightgbm - 0.03063 public LB

Steps

• place train_ver2.csv, test_ver2.csv to ../input/
• install pandas, scikit-learn, numpy, xgboost, lightgbm (or comment out lightgbm part) libs for python3
• set proper number of threads in engines.py
• ./run.sh