Credit Card Overdue Risk Prediction and Optimization Based on Logistic Regression Model
Credit card overdue risk prediction is crucial for risk management of financial institutions.Based on the overdue behavior data of 8 731 credit card users,this paper analyzes seven key applicant characteristics and uses Logistic Regression model to accurately predict the overdue risk.To address the issues of data imbalance and prediction credibility,it uses the One-Hot Encoding technique to process the category data and solves it by sample balancing technique.In terms of model tuning,it takes into account the model Goodness of Fit and complexity synthetically,and fine-tunes the model parameters to significantly improve the reliability of the prediction.Compared with the 75.494%accuracy of the AIC model,the optimized Logistic Regression model shows better prediction performance.In addition,this paper comprehensively evaluates the model performance using ROC curve.The experimental results show that the optimized Logistic Regression model performs well in credit card overdue risk prediction,which provides strong support for risk management and decision-making of financial institutions.
万方数据
维普
