Research On Individual Credit Evaluation Under The Development Of Internet Finance

The rise and development of Internet finance in China presents a variety of financial business models and operating mechanisms,but also the problem of customer credit risk,it is urgent for financial companies to improve personal credit assessment and management to reduce credit risk.Based on a large wealth of customer credit history data,the original data set contains the customer’s personal basic information,personal economic situation,credit factors,personal credit data(overdue type and non-overdue type).Among them,we take the data of personal credit as the target variable,and the basic information of individual,the economic situation of the individual and the credit factor as the characteristic variable to study the personal credit evaluation.Firstly,the personal credit evaluation index system is established based on the existing personal credit research results,and then through the filling and deletion of data missing values and the preprocessing of sample imbalance data,the extracted samples are subdivided into 80%training Set and 20% of the test set,four credit evaluation models of neural network,random forest,AdaBoost,and Logistic are established using the training set samples.The indicators with the feature importance of more than 0.01 are used to evaluate personal credit.The test set is used The data verifies and comprehensively evaluates the four models.Finally,combining F-Score and accuracy,it is found that the F-Score values of AdaBoost and random forest are similar,close to 17%,but the accuracy of random forest model reaches 79.44%,so In this paper,random forest is selected as a model for personal credit evaluation,and 17 characteristics selected by random forest are used to evaluate personal credit.In the analysis of the results,it is concluded that the lower the income level,the more unstable the marital status,the higher the probability of default of features such as unsecured;the status of the credit card account is normal,and the characteristic of the provident fund is lower.The shortcoming of this article is that the recall rate of these four machine learning algorithms is low.Finally,this article gives suggestions for reducing credit risk.The first is to establish a perfect personal credit evaluation system;the second is to sanction the untrustworthy in terms of government and Internet infrastructure,in order to help the healthy development of the P2P industry.