Research On Method Of Estimating Credit Risk For Personal Loans Based On Support Vector Machine

With the rapid growth of China's economy,the scale of the credit loan market is constantly expanding.However,personal credit system is not perfect in China,banks and other financial institutions cannot fully grasp the credit information of the borrowers,and individual loans cannot be managed by the way of enterprise loans.Therefore,it is very important to research the evaluation method of personal credit risk based on the statistical data of personal credit.The problem of borrowers' credit risk assessment can be converted into a binary classification problem of whether the borrowers default or not.Therefore,combining with the classical statistical learning classification method--support vector machine,a new method that is used to evaluate personal loan credit risk is established.Experimental results show that the new method can solve the problem of personal loan credit risk assessment effectively.The main contents are as follows:(1)Firstly,the factors influencing on the credit risk estimation for personal loans and the construction principle of index system are analyzed.Then,three primary indicators,namely personal index,economic index and credit index,are obtained.10 personal indexes,including age,number of credit,debt ratio,monthly income etc,are selected to construct the personal credit risk assessment index system.Secondly,the information value(IV)model is used to calculate the IV value of credit data and show the importance of each feature.At last,the results show that the IV values of available amount ratio,times of 30 to 59 days late,times of overdue 90 days,and times of 60-89 days late are bigger than 0.3,which indicates that these features have more influence on the credit risk assessment of the lender.(2)Firstly,combining with fuzzy mathematics theory,a membership function based on the heterogeneous class hyperplane is proposed,through which different samples are given different weights.Secondly,combining with the traditional support vector machine(SVM)and the proposed membership function,a fuzzy support vector machine(FSVM)based on the heterogeneous class hyperplane is proposed.Thirdly,with the data for personal loan credit risk assessment from Kaggle data set,some comparative experiments are provided.At last,experimental results show that the proposed fuzzy support vector machine based on the heterogeneous class hyperplanecan effectively solve the problem of estimating credit risk for personal loans,which shows the feasibility and the effectiveness of the proposed method.This dissertation proposes an improved statistical learning method,which can not only enrich the theory and application research of support vector machine,but also provide a new method for personal loan credit risk evaluation.