Identification Of Defaulting Customers Of Commercial Banks

With the approaching of the open financial market,the loan of commercial banks in Guizhou province is becoming more and more important to the rapid development of the market economy,but the management mechanism of loan default still needs to be perfected.Because of the imperfect loan risk management system,the lack of scientific credit risk management tools,resulting in the high rate of non-performing loans and the occurrence of credit fraud,it is imminent to improve the management level of credit risk.The key to solve these problems lies in the deep excavation and scientific analysis of credit data.In this context,this paper uses the personal customer credit data of commercial banks to build a default customer recognition model,summarizes the characteristics of default customer behavior,and establishes a credit scoring card to provide the best customer analysis model for default.The main work of this article is summarized as follows:First,preprocessing the personal credit data of Guizhou commercial banks.The data(280 thousand customers with a total of 627 indicators)were preprocessed for desensitization,data cleaning,unbalanced data processing,outlier processing,data discretization,data integration and so on,and the initial data needed for modeling were obtained.Second,using pre processed initial data,we build a variety of classifiers for identifying default customers and compare them.The logistic regression model,the naive Bayes model,the support vector machine model,the decision tree model and the combination model are constructed respectively,which are based on the user behavior characteristics of the commercial banks in Guizhou province.These models are used to predict whether the customer has a major breach of contract,and to evaluate and compare the results of the models.Out of the optimal model.Third,induce the optimal behavior of the default customer.Using the best model to break customer's recognition rules,we conclude important behavioral characteristics.Fourth,establish the credit scoring system.First,based on the Gini entropy decision tree index selection method,116 indexes with important information are selected.Secondly,logistic stepwise regression method is used to determine the index of the final index system.Then,the score value of each index is calculated by the WOE(weight of Evidence)evidence weight conversion method;finally,the index of the construction is given.The standard system is tested for effectiveness.Fifth,customer portrait of the defaulting customer.Based on the behavioral preferences of defaulted customers in various aspects,it vividly reflects the unique characteristics of defaulted customers relative to normal customers.The purpose of this paper is to use the personal credit data of Guizhou commercial bank to reduce the loss caused by customer credit default,and to provide a scientific customer accurate identification method and credit risk management tool.Therefore,this paper not only has a certain theoretical value and application value for the commercial banks to improve the accuracy and scientificity of customer recognition,but also has a reference role in marketing,anti fraud,risk assessment,value added,loss union collection and so on.