Research On Personal Credit Evaluation Method Of Consumer Finance Companies

Under the new economic situation,the traditional financial services are not enough to support the further development of the real economy.From 2010 to 2017,the consumer finance companies in our country are developing rapidly and show great potential.The emergence of consumer finance companies satisfied the credit needs of individual consumers.However,the personal consumer credit business provided by consumer finance companies is characterized by small amount,unsecured,unsecured,simple procedures and rapid approval,which exposes them to great credit risk.The ability to recognize credit risk has become one of the core capabilities of consumer financial institutions.Good risk control can help financial institutions expand their business scale,reduce the loss of bad debts and increase profits.Therefore,establishing a set of accurate and practical personal credit rating system is the key for financial institutions to effectively avoid credit risk and further promote the development of credit industry.Consumer finance personal credit business was started not long in our country,the corresponding financial institutions had a low level of management of credit risk and did not form a systematic personal credit evaluation system.Coupled with the large number of individual customers,machine learning methods have begun to be used in personal credit assessment.Machine learning algorithm is a supervised learning,which can improve the algorithm of empirical data learning to process data.Decision tree is one of the machine learning methods.In this paper,the decision tree and its combined classification algorithm are mainly used to evaluate the personal credit of consumer finance companies.The main research content of this paper is as follows: Firstly,introducing the principle of personal credit evaluation methods.Personal credit evaluation is actually a classification.Correspondly,the choice of the credit evaluation method is the choice of classification method.This paper proposes that when there are too many qualitative independent variables and the level of qualitative variables is too large in the classification data,the evaluation results of decision tree and its combination classification methods are better than those based on quantitative variables such as Logistic regression,support vector machines,etc.Secondly,using the decision tree classification,bagging classification,random forest classification and adaboost classification method to fit the data of this paper,and compare accuracy,risk cost,stability,10-fold cross validation results etc,draw a conclusion that random forest classification method is better than other methods.Again,the above comparison is made between the classification methods of random forest classification,Logistic regression classification,and support vector machine classification.The conclusion that random forest classification still outperforms these classification methods,and confirms that the hypothesis presented in this paper is valid for the data in this paper.The end of the article summarized the conclusion and pointed out the problems that the consumer finance companies in the personal credit evaluation of the specific application process also need to pay attention to.