Research On P2P Lending Network Topology Characteristics And Borrower's Credit Risk Assessment

The development of the Internet has given more possibilities for financial innovation.P2 P online lending,as one of the representative models of financial innovation,is of great significance for connecting social idle capital and inclusive finance.However,problems such as illegal capital injection,unconventional operationand the credit risk of borrowers in P2 P online lending have always been difficult to regulate.Today's regulatory policies are more focused on "exit and transformation".However,P2 P online lending may face to a big change because through the transformation,P2 P lending will develop to a more standardized and healthy financial format.This study of P2 P borrowers' credit risk has reference significance for its future transformation.This article mainly conducts research from the following aspects:First of all,build a complex network of P2 P borrowing-lending relationships,and combine the borrower's(the node's)credit level and default situation to analyze the network's structural characteristics and borrower's behavior from both the overall and partial perspectives of the network.Calculate the input-degree centrality,closeness-centrality,proximity prestige and betweenness centrality of each node in the network,and their correlation with the borrower's credit level and borrowing information corresponding to the node to further explain the practical significance of the borrower's network topology characteristics.Secondly,study the relationship between the borrower's node network topology characteristics and defaults.The results show that 1)the input-degree which is related to the number of borrower's liabilities is positively related to the borrower's default,2)the closenesscentrality,proximity prestige and betweenness centrality which are related to the efficiency of transmission and circulation are negatively correlated with the borrower's default.The results are consistent with the practical interpretation of the borrower's network topology characteristics.Based on the result,,an index system for evaluating the credit risk of P2 P borrowers considering the network topology characteristics is constructed.Finally,using machine learning methods to build a risk assessment model that integrates the borrower's network topology features,and do a empirical evaluation based on the data of Renrendai platform.The empirical result show that the model training and prediction effect is better whentaking into account the network topological characteristics of the borrower and GBDT model is the best.At last,through the calculation of the feature importance of the tree model,it is obtained that the importance of the network topology information is only less than borrower's credit information and loan information.