Because user_pay.txt, user_view.txt and extra_user_view.txt are bigger than 100MB, it can't be pushed onto github, just look at the Get dataset part.
- shop_info.txt: Shop's information.
- user_pay.txt: User's pay history.
- user_view.txt: User's view history.
- extra_user_view.txt:A padding to the original user's view History.
According to Users' history data and shops' information, predict users' pay behavior int he next 2 weeks.
We designed 5 models. the SVM model bring us our top score: 539/4058
.
And the prediction loss on the dev set of every model is:
- LR:
0.00553571428571
- SVM:
0.00107142857143
- KNN:
0.0166071428571
- GBDT:
0.169285714286
- RF:
0.0301785714286