مطالعه مدل اعتبار سنجی اعتبار بر اساس گروه لسو

Credit scoring model is one of common tools for commercial banks to manage credit risks. In this paper, we use a public dataset from UCI machine learning repository and construct credit scoring models based on Group Lasso Logistic Regression, where the tuning parameters λ are selected by the Akaike Information Criterion(AIC), Bayesian Information Criterion(BIC) and Cross Validation prediction errors respectively. The experimental results show that the Group Lasso method is better than backward elimination in both interpretability and prediction accuracy.