DACON Monthly Psychology Propensity Prediction AI Competition

Algorithm | Structured | Classification | psychology | AUC

  • moneyIcon Prize : 100만원+애플워치
  • 2,100 Users Completed

 

private 56등 public 0.78331 (NN)

2020.11.18 14:04 7,015 Views language

NN 모델을 사용했습니다. 
early stopping 안했을 때 public 점수가 더 잘 나오길래 그냥 100 epoch 그대로 돌려서 제출했었는데
오히려 그것 때문에 over fitting이 되어서 실제 private score는 잘 안나왔나봅니다..  혹시나 제 의견과 다른 생각을 가지신 분들 조언 부탁드립니다~!

Code