A Research On Forecasting Personal Loan Defaults Utilizing The Random Forest Model

With the emergence and progression of the personal loan industry,the scale of personal loans has been growing.At the same time,the problem of personal loan default has gradually become the focus of financial institutions.The prediction of whether personal loans will default has become increasingly important.At present,financial institutions mainly use econometric models such as logical regression to predict and evaluate the default risk of lenders.With the advent of large-scale data and the advancements in artificial intelligence technology,some scholars began to compare the emerging machine learning models such as random forest with the linear regression and logical regression models in traditional econometrics,and gradually applied various classification prediction models to the research of personal loan default prediction.Therefore,it is expected to have certain research significance and application value to apply the stochastic forest model to personal loan default prediction and analyze the influencing factors of personal loan default.The structure of this thesis is specifically outlined as follows: First,it analyzes the current development status of personal loans at home and abroad,as well as the traditional organizational structure,indicator system and prediction methods of personal loan default prediction,and summarizes the problems in the current personal loan default prediction.After analyzing the characteristics of stochastic forest model and its advantages in personal loan default prediction,this thesis discusses the specific use scenarios and relevant prediction and evaluation indicators of personal loan default prediction,and analyzes the applicability of stochastic forest model in personal loan default prediction.Finally,the stochastic forest model is used to carry out empirical research on personal loan default prediction.Through the personal loan data information of Lending Club platform,the stochastic forest model is used to predict personal loan default,and the prediction performance of the stochastic forest model is improved by adjusting parameters and optimizing the model;Then,through the comparison and analysis of the prediction results with the logistic regression and decision tree models,the advantages of the stochastic forest model in the loan default prediction are verified;Finally,the importance ranking of the characteristic variables of the stochastic forest model in the prediction of personal loan default is output,and the influencing factors of personal loan default are analyzed.Ultimately,based on an assessment of the current pertinent status and findings derived from empirical inquiry,this manuscript consolidates the entirety of its contents,proposes corresponding remedial measures and provides recommendations accordingly,and this thesis anticipates future research opportunities.The research findings and deductions of this study are summarized as follows:(1)The current indicator system and prediction method for personal loan default prediction are insufficient,The indicator system is mainly static indicators,and the forecasting method does not fully analyze and utilize the personal loan data information.(2)Applying stochastic forest model to personal loan default prediction can achieve better prediction results.Even the unoptimized random forest model has better prediction effect than the logistic regression and decision tree model.And by optimizing the parameters of the stochastic forest model,it can further improve the prediction performance of the model for personal loan default.(3)The random forest model can output the ranking of the characteristics and importance of the model when forecasting,give the weight of different indicators on the prediction results of personal loan default,and finally conclude that the main factors affecting personal loan default are credit score,current loan amount,debt level,income and employment status.