The Application Of Decision Tree Method In The Classification Of Bank Borrowers

With the development of the times and the economy,people's consumption concept is gradually changing.Bank loan is an important form of consumption in advance,the use of loans for borrowers to cash flow,used to meet the purchase,buy a car,as well as studying abroad and other current consumer demand.Because of the high interest on loans,the banking industry is willing to develop the loan business to the demand for loans,from which to obtain greater profits.However,the borrower default will also cause greater losses to the bank,the bank borrowers to classify,identify potential breach of contract,refused to apply for loans,will reduce the risk of bank losses.This paper first introduces several commonly used decision tree algorithms,including ID3,C4.5 and CART algorithm,and combined model of decision tree algorithm-random forest,with analyzing the relationship and difference between various algorithms.Subsequently,the bank loan data for a detailed description,which can be divided into family,work income and credit card loans three plates.Then,C5.0,CART and random forest algorithm were used to model the data,and the prediction accuracy,true positive rate and true negative rate were used to evaluate the model.In the meantime,the cost matrix and ten fold cross validation are introduced to adjust the parameters of the model,and the effect of the model is evaluated by using Kappa statistics and ROC curve.Finally,the results of the three models are compared,and the results show that the effect of C5.0 and random forest are quite different from that of CART.The reason lies in that each model is balance of three correct rate:the prediction accuracy,the true positive rate and the true negative rate.Since there is no absolute good or.bad,you need to select the appropriate model according to the actual requirements of the bank.