Research On Internet Financial Risk-Control Evaluation Based On Data Mining

Internet finance(ITFIN)is defined as an emerging technology that dynamically integrates Internet technology and financial functions.Generally,ITFIN can be divided into six categories,including third-party payment,online debit and credit,big data finance,crowdfunding,information-based financial organizations and INFIN portals,while the category of online debit and credit include online P2P loan and online petty loans.For online P2P loan,risk control is classified into seven groups,including risk control before examination and approval,risk control during examination and approval,risk control during credit extension,risk control during post-loan stock customer management,risk control during post-loan management of customers with overdue behaviors,risk control during financial liquidity management,and risk control during lending,while the first three are the most important parts and can be called by a joint name pre-loan risk control.The pre-loan risk control in online P2P debit and credit is the emphasis and orientation of the discussion in this paper.For ITFIN lending enterprises,the essence of risk control is to model user data and price the risk.Mostly,a risk pricing model is based on the manner of rating users'credit scores.Banks usually adopt the credit investigation reports issued by the People's Bank of China(PBC)to uniformly rate customers' credits,and the rating results are taken by many banks as one of the important considerations for pre-loan risk control.For ITFIN lending enterprises,the PBC rating results are good,but the PBC's credit investigation reports only provide a simple version of the evaluation report to outsiders,and only the banking institutions have has the right to view the full version of the credit reports.This kind of brief credit investigation report is not enough for ITFIN lending companies.This market vacancy has also contributed to the rise of third-party credit investigation agencies.Although the PBC's credit investigation data is not open to third-party agencies and ITFIN lending companies,they can rely on the advantages of the Internet,and complete the pricing of a user's loan risk based on the collected massive user data,consumer behavior data and social behavior data.The value is similar to that of the PBC's credit investigation in financial scenarios,and,even in some specific scenarios,is superior to the PBC's credit investigation in risk control.The most representative cases are the Qian Hai Zheng Xin(credit investigation)and the Ant Financial Service Group.These two Internet financial technology companies are also large-scale enterprises in the ITFIN field.The Ant Financial Services Group focuses on users' e-commerce scenarios,and has most of its data sourced from e-commerce business scenarios such as Taobao,Tmall and Alipay.The data of the Qian Hai Zheng Xin is mostly from the banks subordinated to PINGAN Group,the insurances and other scenarios.The difference between the two is related with the ecological features of their products.The data of Qianhai Credit Information is more from the bank,insurance and other scenarios of Ping An Group.This difference is related to the ecological characteristics of its products.The growth of the post-90s consumers who are bold in premature consumption is accompanied by the booming of credit ITFIN.The ITFIN credit companies have launched a brand new risk control model to quickly fill the gap in the market,analyzed and model massive user data with data mining technology,constructed a new risk control and evaluation mechanism,conducted technology and algorithm-driven analysis of massive data for precise user portraits,completed pricing of risk control which is beyond the capacity of traditional banks or cannot be achieved with a low cost,and therefore reshaped the pricing of financial risk.This route,for the one hand,reduces the cost of risk control and upgrades consumer satisfaction,and,for the other hand,significantly improves technical indexes(such as the efficiency of risk control and evaluation,and the peak capacity)over those of the traditional models.For one thing,the new risk control model is based on automatic computer review,and requires no artificial intervention;for another,the borrowing risk of a user is rated by observing some indexes,and this model significantly reduces the cost of examination and approval,and increases enterprises'profits.Firstly,the background,significance and objectives of pre-loan risk control and evaluation in the ITFIN industry and related research methods are expounded in the paper.Besides,relevant theoretical foundations,technologies and knowledges involving risk control and evaluation are introduced with emphases on behavioral finance,data mining-related knowledge,and the decisionSecondly,the classifications of the indexes involved in ITFIN lending enterprise for a risk control and evaluation model and the principle for the model construction are then explained in the paper.During the analysis,the risk metrics and psychological decision-making knowledge of behavioral finance are utilized to divide all data indexes into four major categories--authentication,biological cognition,loan repayment and behavioral data.On this basis,a total of 47 risk control indexes are selected as the indicators for the risk control and evaluation model discussed in this paper.Thirdly,how to design the risk control and evaluation data warehouse according to the data warehouse star-topology specification is discussed in the paper with detailed elaboration on the modeling practice principles and specific field meanings of the data warehouse.In addition,the form of activity diagram is used to display the detailed data warehouse's clean-convert-load(ETL)steps.Then,the Weka tool is used to analyze and verify the data in the data warehouse with the C4.5 decision tree algorithm,obtaining the experimental conclusion of an accuracy rate of 96.49%.Lastly,the existing problems and imperfections in the experiment of the risk control and evaluation are analyzed and summarized,while the future research orientations are prospected.The paper is innovative in the behavioral financial theory-based analysis and modeling of borrowers' behavior data,hoping to find out the relationships between the behavior data and the risk control and evaluation model.