Research On Personal Credit Evaluation Model Based On Lasso-logistic And Neural Network

With the acceleration of China's financial system reform and the improvement of residents' income level,the proportion of consumer loans in the financial business has been increasing.The consumption mode of credit economy has been widely recognized in people's daily life.Financial institutions have gradually shifted their focus of operation to consumer credit business.The development of credit economy brings great convenience to social life,but at the same time,the increasing scale of credit also brings huge credit risks to financial institutions.Therefore,it is of unprecedented importance for both traditional financial institutions and emerging internet financial enterprises to evaluate the credit status of borrowers scientifically and comprehensively.This is not only the main problem of academic research,but also the strategic problem that major financial institutions need to solve urgently.More and more researchers and business departments are investing in the field of personal credit assessment,aiming at exploring more scientific and efficient algorithms to achieve user credit assessment.This paper not only realizes the high accuracy of the evaluation,but also considers the robustness and interpretability of the model comprehensively,aiming at obtaining a better personal credit evaluation model in all aspects.Logistic model based on Lasso theory can not only select important characteristic variables,but also has higher prediction accuracy than full-variable Logistic model and stepwise regression model.Aiming at the problem that BP neural network is easy to cause non-convergence and time-consuming,parameter optimization is carried out in the process of model building to achieve the best prediction effect.Then Lasso-logistic regression model and BP neural network model with optimized parameters are combined to establish the combined model.The characteristic variables selected based on Lasso theory and the output of Lasso-logistic regression is used as input variables of BP neural networkmodel.After defining the concept of personal credit evaluation,this paper constructs the index system of personal credit evaluation based on domestic and foreign experience.The open credit data set is selected to validate the personal credit evaluation model in this paper.Before establishing the model,the sample data are qualitatively analyzed,which provides a relatively intuitive basis for the interpretability of the personal credit evaluation system.Then the data preparation work is completed,including index assignment and standardization processing.The data set is divided into training set and test set.The data preparation work is done for the establishment of personal credit evaluation model based on Lasso-logistic regression and neural network.Experiments show that compared with Lasso-logistic regression model and BP neural network model,the combined model proposed in this paper takes into account the characteristics of high accuracy,strong robustness and better interpretability.Finally,the paper summarizes and looks forward to the work.