Construction Of Customer Signature And Prediction For The Loss Of Bank's Personal Loan Customers

With the reform of China's socialist market economic system,competition in the financial market has become increasingly fierce.The business model of bank has been gradually updated and improved,which has led to a gradual decrease in the level of personal loan business in banks.The involvement of small loan companies and loan apps has led to the loss of individual financial services in the financial market.Nowadays,prediction of personal loan customers and customer retention have become an important research topic,which is crucial for the future development of the bank.In this paper,the real data of 80000 individual loan customers in commercial banks are selected for modeling and analysis.In the construction of the customer signature,the training data set and test data set,variable selection,data normalization,variable clustering and factor rotation are firstly constructed,and then the K-means clustering method is used to construct the customer signature.The results of clustering out six categories are obtained,and the attribute characteristics of each category are described.Finally,the corresponding retention strategy recommendations are proposed.Variable selection,oversampling,correlation testing and variable clustering steps are required before predictive modeling.After obtaining the prediction model,the test data set is used for judging the model based on ROC curve.Assume that all data is taken before the customers settle the loan.The observation window period is 6 months,and the criterion for the loss is that the variable bad_good is 1 which means that the personal loan customers are loss.In this paper,80000 samples and 625 variables are used for cleansing,and then processed for dimensionality reduction,and ultimately reduced to 16 variables for modeling.Then,we use the methods of logistic regression,decision tree and random forest to model the dimension-reduced data,and compare the results of the model,find out a good and stable prediction model.This paper systematically classifies bank's personal loan customers,constructs customer signature and prediction for them,which can effectively classify personal loan customers,effectively assist banks to prevent the loss of personal loan customers,and improve ability of business to compete to help the bank's future development.