# My simple trick for this competition

train['final'] = train['bin_predict']*(-33.21928)+(1-train['bin_predict'])*train['no_outlier']
0.015 boost in local cv compare with same feature train directly
I think this trick works because binary is better than regression even though metric is rmse when label is 1&0
like Kaza's post
[https://www.kaggle.com/c/avito-demand-prediction/discussion/59885](https://www.kaggle.com/c/avito-demand-prediction/discussion/59885)
this competition is rmse too,but Kaza use xentropy as their objective
add：
Classifier AUC: 0.914 
no outlier Regression CV:1.545
the most important features for binary_predict from this post
[https://www.kaggle.com/c/home-credit-default-risk/discussion/64503#378162](https://www.kaggle.com/c/home-credit-default-risk/discussion/64503#378162)