In this paper,we use three data mining techniques and advanced statistical software SAS to build application credit scorecard models,based on the customer data of a commercial bank.The results showed that three models have some prediction ability,but logistic regression model is the best one and is worth applying in banks.Data is the foundation of building credit scorecards.But in practical it usually was polluted.So it is necessary to address data before building credit scorecards including data cleasing,data conversion,and derivation of variables.The innovation of the thesis lies in:1.Using the decision tree,logistic regression and support vector machine to set up scoring models and make a prediction;2.We make a comprehensive comparison of the prediction results from the aspects of model's precision,stability,applicability and interpretation and get the general conclusion. |