Actions on the basis of the account according to the risks of information, information and transaction behavior, through data analysis, from the account features, equipment operation and transaction frequency, trade time, trade amount, regional distribution characteristics of dimensions such as mining, supervised machine learning model, effectively identify high-risk trading account.
train with label and test without label.
PCA,Feature Engineering,Feature importance extraction based on classification decision tree,xgboost,bagging and Pearson correlation matrix.
out-bagging.csv and out-xgboost.csv for the predict of credict risk for the users in test-without-label data.