Research On Credit Risk Assessment Of Peer To Peer Lending Based On SVM

In recent years,the Internet financial sector in China has received rapid development.P2 P net lending is an important aspect both in financial innovation and Internet Banking.The development of P2 P net lending has played a positive role in broadening investment and financing channels and in promoting market consumption.However,defaults happened from time to time in P2 P net lending platforms,as establishment of domestic individual credit system is imperfect,operation time of P2 P net loan platforms is relatively short,collection of personal credit data is very limited,asymmetric information between the lender and borrower is large enough,which has seriously influenced healthy development of P2 P net lending.It is of great theoretical and practical value to establish effective credit evaluation system to borrowers,to adopt effective means of credit risk assessment,and to avoid or reduce the occurrence of credit risk from borrowers.Scholars at home and abroad have made in-depth researches on characteristics of P2 P net lending borrowers,establishment of credit risk system and feasible risk assessment methods.Both traditional statistical methods and AI have been applied to credit risk assessment of net lending.Support Vector Machine(SVM)theory is a novel machine learning method featuring model building with fewer samples,high-efficient operation and high-accurate classification result,which has showed good application prospects and unique edges in the field of credit risk assessment for P2 P net lending and has become a hotspot of the study on credit-risk evaluation method.The research has trained and optimized net lending credit risk evaluation model based on SVM by analyzing and processing loan data from domestic leading P2 P net lending platform(renrendai.com),establishing credit risk evaluation system in line with net lending characteristics in accordance with scientific and effective principle and using libsvm-3.22 software package to program procedures.The research has demonstrated that SVM is an effective method for net lending credit risk evaluation by comparing the results of classification of different kernel functions,using heuristic method to adjust penalty coefficient C and gamma value based on SVM model with Radial Basis Function(RBF)kernel,and optimizing accuracy of model in data prediction.The research results showed that SVM model adjusted has received sound classification accuracy for net lending data,up to 93.46% and has gained stable classification effect.The paper made an analysis and discussion in classification effect and parameter selection of SVM model kernel function,and put forward policy recommendations aiming at strengthening P2 P net lending platform management and reducing its credit risk.