Big Contest Championship league

Team Da dejus @lhmlhm1111 @codenavy94 @dockjong @Hail-cali

Baseball ops prediction

Code description

• how to train?
  • we worked code based on shell command with python for faster hyper parameter tuning
  • the major options you should input, are -models --y_feature, --file, --data_path
  • for -models, -l option is model list you want to train, it is mapped inside code with dict

python final_train.py -l xgb lgbm rf --data_path  ./dataset --file train.csv --y_feature 장타

• hyper parameter tuning

python run_grid.py -l xgb lgbm rf --data_path  ./dataset --file train.csv --y_feature 장타

• gird_search & inference

python run_grid_stack.py -l xgb lgbm rf --data_path  ./dataset --file train.csv --y_feature 출루

---

serving dataset

• the major option for making dataset from raw is ws (windeow size) and agg(aggregation rule |date | game|)

modeling

• we used stacking ensemble model