The Application Of Combined Model Based On LASSO-SVM And Logistic In Personal Credit Evaluation

In the era of big data,China's financial industry is developing rapidly,and the people's consumption concepts and investment methods are constantly changing with the times.People have changed from the original concept of savings to advanced consumption,which has prompted China's credit industry to flourish.At the same time,the issue of personal credit risk has become increasingly prominent,posing a huge challenge to the credit industry.The personal credit evaluation model is based on data mining technology to objectively predict the default risk of loan customers.In western developed countries,from the early statistical methods to artificial intelligence methods,the research and application of related technologies have become mature.Due to the late establishment of the personal credit information system in China,the personal credit system is not perfect,and related research is relatively backward.In this context,research on how to construct a personal credit evaluation model with higher prediction accuracy and how to construct a scientific indicator system is of great significance to the development of China's credit industry.This paper takes the personal credit evaluation combination model as the research object.First,a literature review of the research on personal credit evaluation models is conducted.Then the Logistic model principle,LASSO method,SVM model principle and combination method are introduced.The problem that SVM model can't filter variables is combined.LASSO method and SVM model are combined to construct LASSO-SVM model for Logistic model and LASSO.The SVM model has a limited prediction accuracy,and a logistic and LASSO-SVM combination model based on least squares is constructed.Secondly,it summarizes and analyzes the index system commonly used by domestic scholars,and establishes the personal credit evaluation index of this paper based on the principle of personal credit index system construction and the credit data set of this paper.Finally,the German credit data set on UCI is used as the empirical data set.The original data is preprocessed,and then the Logistic model,LASSO-SVM model and Logistic and LASSO-SVM combination model based on least squares method are established.The performance of the LASSO SVM model is affected by the parameters and is optimized by grid search.The confusion matrix and ROC curve are selected as the model evaluation criteria,and the output results of the three models are compared.The results show that the total classification accuracy and AUC value of the Logistic and LASSO-SVM combination models based on the least squares method are better than the other two single models.High,the LASSO-SVM model has a higher overall classification accuracy and AUC than Logistic.From the first classification error rate and the second classification error rate,the logistic and LASSO-SVM combination model based on the least squares method is lower than the other two single models.The above empirical results show that the Logistic and LASSO-SVM combination model based on least squares method can be applied to personal credit evaluation.