时间节点:2018年10月22日,00:00:01-2018年10月24日,23:59:59【复赛入围审核】
@anxiangSir,@irene9adler,@maomao1994
-(1)xgboost(Try TM
-(2)catboost(Try TM)
-(3)tidy-xgb(Try TM
-(4)SVC+RF+LGBM+XGB(Try TM)
-(5)
-(6)wide and deep(Try TM)
-(7)lgb(OK AX)
-(8)RF(random forest)(Try SQG)
-(9)gbdt+LR (Logistic regression)(Try SQG)
-(10)FTRL(https://www.kaggle.com/c/mercari-price-suggestion-challenge/discussion/47295)(https://www.kaggle.com/leeyun/ensemble-model)(Try SQG)
-(11)NFM(Factorisation machine)(OK AX)
-(11)深度FFM(Try AX)
-(13)DeepFM(Try AX)
-(14)FNN(Factorisation machine supported neural network)(Try AX)
-(15)CCPM(Convolutional click prediction model)(Try AX)
-(16)PNN-I(Inner product neural network)(Try AX)
-(17)PNN-II(Outer product neural network)(Try AX)
-(18)PNN-III(Inner&outer product ensembled neural network)(Try AX)
-(1)用xgboost来stacking训练lightGBM(Model1)
-(2)用lightGBM来stacking训练xgboost(Model2)
-(3)异常变量的处理
-(4)为什么进行特征选择?单独的特征没有用,组合的特征可能有用?
-(5)分布不对称:取对数试试?
-(1)https://blog.csdn.net/xiewenbo/article/details/52038493
-(2)xgboost:http://www.52cs.org/?p=429
-(3)https://github.com/guoday/Tencent2018_Lookalike_Rank7th