Probability Estimation Via Weighted Support Vector Machine In Personal Credit Scoring

In recent years,with the rapid development of China's economy,the business of personal micro-credit continues to grow,and the demand for personal consumer loans,such as housing mortgage,credit cards and auto loans,gradually expands.In a credit loan,the applicant?client?is usually concerned about making the loan on time.Financial institutions,on the other hand,worry about whether the applicant?client?can repay the loan within the predetermined time,in order to reduce the financial loss caused by credit default.At present,China's personal credit evaluation system is not perfect,loan default occurs from time to time.In order to reduce losses,save costs and make scientific loans,an accurate and reasonable personal credit evaluation system is needed as a support.The core of the personal credit evaluation system is to establish a credit evaluation model based on the personal information of the previous credit applicants,and predict whether the new credit applicants can repay the loan on time according to the personal credit evaluation model,so as to decide whether to grant loans for them.A good credit evaluation model can not only help financial institutions to accurately predict the applicant's personal credit,so as to accurately and efficiently handle loan approval,effectively avoid credit risks and reduce economic losses,but also help applicants with good credit to apply for loans more quickly and improve economic efficiency.By reviewing past personal credit evaluation model,this paper found that the current personal credit evaluation model problems are as follows:first,modern society has been in the age of big data,personal credit evaluation data not only large amount of data,and the data dimension is higher,the traditional personal credit evaluation model based on statistical learning methods calculate the cost;Second,although the traditional personal credit evaluation model based on machine learning is fast in calculation,most of it can only be classified and cannot be used for probability estimation.If the personal credit evaluation model can be used for probability estimation,the applicable scope of the model can be greatly improved.In view of the above shortcomings,this paper improves the traditional support vector machine?SVM?model,establishes the probability estimation model of weighted support vector machine?w-SVM?,and applies the probability estimation model of w-SVM in personal credit evaluation.First,the w-SVM probability estimation model weighted the loss function of two types of samples,namely"defaulting"client and"non-defaulting"client.Secondly,in the credit approval process,different amounts of loans have different requirements for users'reputation.The probability estimation model of w-SVM can estimate the probability of two types of samples,set the threshold according to the actual situation,and finally classify customers into"default"client and"non-default"client,so as to improve the applicability of the model.Finally,this paper compared the probability estimation model of w-SVM with logistic regression model,random forest model and traditional SVM model,and comprehensively considered the accuracy,recall rate,precision,F₂ and area under curve?AUC?indicators of the model prediction,confirming that the probability estimation model of w-SVM has significant advantages in personal credit evaluation.