Credit Rating Of Online Lending Borrower Using Recovery Rates

Online lending is a new Internet financial model.Since 2013,the online lending industry has been developing rapidly in China.However,with the continuous expansion of the industry scale,various problems are gradually exposed.High proportion of problem platforms and frequent thunderstorm events have seriously harmed the healthy and sustainable development of the industry,so it is urgent to measure credit risk reasonably.Among them,the key problem is how to establish a reasonable credit rating mechanism for borrowers.As an important factor in the credit risk model,recovery rate1 is of great research significance(BIS,2011).A reasonable credit rating mechanism is required to effectively reflect the borrower's default recovery rate(Chi et al.,2019;Chen et al.,2019).At present,there is not a unified borrower credit rating mechanism in the online lending industry,and there are great differences between the credit rating mechanisms of different online lending platforms.At the same time,few of the borrower credit rating mechanisms take the recovery rate into account.Therefore,taking Renrendai(https://www.renrendai.com/)as an example,this paper uses the recovery rate to evaluate the effectiveness of the existing online lending platform borrower credit rating mechanism,and then gradually establishes a factor score K-means clustering online lending borrower credit rating mechanism based on the recovery rate,aiming to provide theory support for the credit risk management of online lending.Based on the above theory,the analysis and construction of borrowers' credit rating in this paper are mainly divided into three steps.The first step is to collect the sample data of 14558 personal lending default assets from 2011 to 2016,and divide the data into seven categories(AA,A,B,C,D,E,HR)according to the credit rating of personal lending to borrowers.Using the boundary kernel method and Kruskal-Wallis test,this paper finds that Renrendai's credit rating mechanism can't distinguish the borrower's credit from the recovery rate,that is,the difference of the borrower's recovery rate of different credit ratings is not obvious,which is not consistent with the importance of the recovery rate in credit rating mentioned above.Second,considering that the main reason why Renrendai's credit rating can't distinguish the borrower's credit is the lack of reasonable characterization of the borrower's characteristics,this paper uses the most commonly used K-means clustering method in the current credit rating literature,and refers to two typical borrower's credit rating models(FICO credit rating system and sesame credit rating system),15 indicators are selected,including 7 qualitative indicators(education background,marriage,income,company size,working years,real estate and real estate loans,car loans),7 quantitative indicators(amount,interest rate,term,repayment ratio,serious overdue ratio,certification information,regional per capita disposable income)and 1 interval indicator(age).After completing the selection of credit rating indicators,this paper standardizes the indicators and preprocesses the above-mentioned data of Renrendai borrowers.A total of 13467 groups of data are obtained.Finally,K-means clustering algorithm is used to cluster 15 indicators for 100 times,and the evaluation indicator of clustering quality-silhouette coefficient is used to compare the quality of each clustering operation,so as to select the best clustering result.At the same time,K-means clustering can only realize the classification of data in the existing research on the credit rating of online lending borrowers,but it lacks a unified standard when judging the credit rating of each category.As the recovery rate is one of the three pillars of credit risk measurement(BIS,2011;Chen et al.,2019),this paper introduces the recovery rate as the standard to determine the credit rating.After re-rating the data,in order to test whether the credit rating method established in this paper is reasonable,this paper uses the boundary kernel method and Kruskal-Wallis test again,and proves that there are significant differences in the borrower's recovery rate among the credit ratings obtained by K-means clustering analysis,so as to solve the problem that Renrendai's credit rating method can not distinguish the borrower's recovery rate.However,there is still a major problem in the credit rating system obtained in this step,that is,under the K-means clustering credit rating method based on recovery rate in this paper,the number of borrowers with each credit rating presents an "inverted pyramid"(borrowers with high credit rating are more than borrowers with low credit rating),which is inconsistent with the fact that borrowers with high credit rating should be less than those with low credit rating.The third step,based on this,this paper analyzes that the main reason for the "inverted pyramid" type is that there may be correlation between 15 indicators and more noise.Therefore,this paper uses factor analysis to reduce the dimension of indicators,and obtains 6 factors(historical credit status factor,repayment ability factor,asset status factor,working status factor,macro environment factor,online lending product factor)that reflect the credit status of online lending borrowers.On this basis,K-means clustering algorithm is used to cluster 100 times,and silhouette coefficient is used to compare each clustering quality,so as to select the best clustering results.For the results of K-means clustering after data re-classification,this paper introduces the default recovery rate as the standard to determine the level of credit rating.The experimental results show that the K-means cluster credit rating method based on the factor score of recovery rate in this paper also plays a role in distinguishing borrowers from the perspective of recovery rate,which is better than Renrendai credit rating method.At the same time,under the factor score K-means clustering credit rating method based on recovery rate,the number of borrowers presents a "pyramid" type,that is,the higher the credit rating,the fewer the number of borrowers,which solves the "inverted pyramid" problem of K-means clustering credit rating method based on recovery rate.To sum up,the biggest innovation of this paper is to introduce the recovery rate into the credit rating mechanism as a quantitative standard to determine the level of credit rating.At the same time,this paper proposes an effective credit rating method,that is,the factor score K-means credit rating method based on recovery rate.