| The network lending business model has special inclusive financial value,fills the gap of the traditional financial service model,improves the efficiency of the financial industry serving the real economy,and solves the financing needs of many small and micro enterprises,farmers and other financing difficulties groups,and expands the financial the investment boundary of market investors.However,faced with more serious information asymmetry,the frequent occurrence of illegal,fraudulent and other thunderstorms in the network platform has made it the “most dangerous” financial sector for investors to look forward to.According to statistics,only 20,000 cases of illegal fund-raising have been established in 2018,involving a total amount of 300 billion yuan.In this context,market participants rationally analyze the effective information of the online lending market,and warn the problems existing in the online lending market,which is related to the future development of the platform and the strategic goal of achieving inclusive finance.The borrowing text language in the online lending market reflects the investor's willingness to repay and the ability to repay.Different linguistic features affect the reader's reading perception,which is a crucial factor affecting the decision-making behavior,and becomes an important information to judge the borrower's default risk..Firstly,based on the literature research,this paper focuses on revealing the influence of the borrower's willingness to repay the textual language characteristics on the decision-making of investors and its role in credit risk prediction.The rooted theory is used to divide the borrowing text language into three characteristics: complexity.,enthusiasm and deception,put forward four research hypotheses.Then,the Renren loan platform is selected as the empirical analysis platform,and the effective financing data of the platform from 2010 to 296,188 is taken as the research sample.The text recognition and feature extraction methods are used to quantify the language characteristics of the borrowing text.Finally,under the condition of controlling other influencing factors,the econometric regression model of the influence of textual language features on financing status is established,and the credit risk prediction model based on Random Foreast,XGBoost and LightGBM classification algorithms is constructed to conduct empirical analysis and test of large sample data..The research results show that 1)the borrowing text language feature brings additional information to investors in the online lending platform,which eliminates the uncertainty of whether the investor lends funds and the borrower defaults to some extent;2)the text of the borrowing text Different characteristics affect investors' investment decisions.The lower the complexity of the loan text,the higher the enthusiasm,the lower the deception,the higher the financing success rate,the higher the financing amount and the lower financing interest rate.Paying attention to the influence of subjective borrowing text language characteristics on investors' investment decisions can help improve the information transparency of online lending market;3)The addition of different characteristics of borrowing text language improves the recall rate and AUC of credit default risk prediction,among which LightGBM The AUC of the algorithm reaches 85.84%,showing a better early warning capability.Based on the above research conclusions,combined with the current P2 P development status and problems,it proposes countermeasures from the aspects of information review and platform services,risk prevention and risk warning,platform supervision and financial innovation,in order to help investors in the online lending market to identify effective information.Scientifically assess the risk of personal credit default and promote the online lending platform to “turn the crisis back,return to equilibrium,and develop healthily”. |
PDF Full Text Request