Prediction Model Of Repayment Probability Of Small Loan Delinquent Customers

With the renewal of China's economic development and the concept of residents' consumption,microfinance institutions have achieved rapid development both in terms of quantity and scale.As a low-risk and easy-to-operate consumer-friendly borrowing platform,the microfinance institutions' income in recent years has also been rather lucrative,but they also bear the risk that some loans cannot be recovered.How to effectively assess and control this risk is an important issue faced by financial institutions such as small loans.With the increase in the overdue ratio and the number of bad debts,the risks of the microfinance institutions are becoming more prominent.Therefore,how to better judge whether overdue debtors will make repayments is an urgent problem for all financial services industries.In order to make more efficient and intelligent judgments,it provides decision-makers with effective decision-making support,thereby improving the accuracy and efficiency of the financial services industry in assessing the creditworthiness of lenders in the approval process.In this study,the logistic regression model and the Naive Bayesian model were used to predict the probability of overdue debtor repayment.This paper chooses forecasting models and variables according to local conditions on the basis of foreign references.In order to detect the accuracy of the model,the sample was divided into training set samples and test set samples according to the ratio of 7:3 before modeling.Then,overdue customer repayment probability was constructed by using logistic regression and naive Bayes method for training set samples.For the prediction model,the accuracy of the model's prediction results is tested through the test set samples.In addition,the clustering analysis method is used to divide the sample set into three categories,and a repayment forecasting model is constructed by using the logistic regression method for these three categories of samples.The results of the sample characteristics and models obtained before the final classification and after the classification summarize the characteristics of the overdue debtors and the stability and accuracy of the models obtained by the two modeling methods are compared and analyzed.