P2P Personal Credit Risk Identification Program Planning Incorporating Soft Information

P2P network lending is the product of the combination of the financial industry and the Internet.As a new form of lending and financing,the explosive growth in recent years has also attracted widespread attention from all walks of life.And credit risk is one of the most common and basic risks in P2 P network loan industry.Whether it is personal credit risk caused by the borrower's own default or platform credit risk caused by P2 P network loan platform running,it will eventually cause economic losses to the lender.The traditional P2 P network loan personal risk evaluation mainly uses the expert scoring method,which is difficult to accurately identify the borrower's credit risk based on the rule-based method,while some P2 P network loan platforms try to use the logic regression,support vector machine and other supervision learning methods,the effect is relatively dependent on the quality of data,the amount of data is less and the data imbalance and other problems will affect the results of risk evaluation Ring.At present,the catboost model of integrated learning method has been well applied in several fields.At the same time,the soft information is also concerned in the field of risk control due to its rich information features.However,the current soft information mainly applies income certification and other applications in credit risk identification.The research of text soft information is less,while the text soft information contains more personal credit risk Information,so the use of text soft information is a hot issue.In order to build an effective P2 P personal credit risk identification scheme,this paper combines soft information and catboost model to build a credit risk identification scheme,and optimizes the parameters of the model based on genetic algorithm to improve the ability of P2 P platform credit risk identification.In this paper,P2P personal credit risk prediction is studied.Firstly,based on the relevant theory and literature,the influencing factors of P2 P personal credit risk are studied.In this paper,the influencing factors are divided into hard information and soft information,and the hard information index system and soft information index system are constructed respectively.Secondly,this paper usesLDA theme model to process soft information and form soft information index.Third,this paper combines soft information and catboost model to build a P2 P personal credit risk prediction model;fourth,this paper uses the empirical samples of Renren loan to compare and analyze the support vector machine(SVM),catboost and lightgbm models,which verifies the effectiveness of the catboost model.The conclusions of this paper are as follows:(1)on the basis of relevant theories,this paper designs 20 hard information indicators,uses loan description information as soft information,and uses LDA theme model to form soft information indicators;(2)through the construction of credit risk prediction model integrated with soft information and catboost,we can see from the samples of renren.com that each model evaluation index By the way,catboost model has the best performance.(3)Through the analysis of the importance variable score of catboost model,we can see that the subject index formed by soft information has an important impact on the classification effect of the model and is of great significance to improve the prediction performance of the model.The research of this paper has certain reference value for further improving the performance of P2 P credit risk prediction,and provides certain reference for promoting P2 P platform to improve its own credit risk management ability.