Forecast Of Overdue Credit Of Commercial Banks

With the development of society,the scale of our national credit market has expanded dramatically,and the development prospects of commercial banks have presented a new atmosphere.At the same time,preventing personal credit fraud and reducing credit non-performing rates have become an important issue for commercial banks to study.The scale of our national credit market has expanded dramatically,but the quality of the matching business has not been greatly improved.Especially in recent years,commercial banks have experienced a series of problems,such as overdue loans and non-performing loans.The problems that have caused huge losses to the commercial banks.Moreover,with the expansion of financial risks in commercial banks,the potential financial crisis has increased,which has had a great impact on the sustainability of China's financial industry.Therefore,it is a necessary question to predict whether a commercial bank customer will have a credit overdue situation.According to the status quo of commercial bank credit,this paper uses the customer credit data provided by a bank to analyze the original data and finds that the data has category imbalance,some variables have a lot of missing values and outliers,and the original variables are provided.More,for these problems,this article deletes the variables with a lot of missing data,and then uses the decision tree and linear regression to complete the missing values of the remaining variables based on the different forms of the data based on the outliers..For the problem of data category imbalance,the original data is reconstructed by SMOTE algorithm to ensure that the reconstructed data basically meets the same number of samples in different categories.After processing the data,this paper uses the IV algorithm to screen the original variables,and then combines the ROC analysis with the selected variables to establish a logistic regression model to test the feasibility of establishing a predictive model for individual variables,and then combine ROC analysis with random forests.Since different variables play different roles in the model,variables with a cumulative importance of 50% are selected and combined to establish a logistic regression model.Finally,the one with the highest AUC value is taken as the variable for establishing the credit overdue prediction model.Finally,the established logistic model is used to predict the credit data of commercial banks,and the prediction accuracy is over 70%.At the same time,the model is compared with the SVM model and the RF(random forest)model.According to the comparison results,the credit overdue forecast established in this paper is explained.The model has certain credibility,and its forecast results can be used as a reliable basis for commercial banks to make decisions,improve their prevention of credit risk,and provide a theoretical basis for financial institutions to formulate correct credit loan policies.