Research On Network Credit Anti Fraud Technology Based On Machine Learning

With the rapid development of the information age,the thinking of the Internet +is being widely disseminated,reaching all walks of life in society,and the Internet has brought tremendous changes to the financial industry.On this basis,the P2 P network lending platform that is born also faces challenges from the initial Flourish to intense competition.Peer-to-peer is a platform for personal lending to customers.The loans are generally unsecured loans and do not require financial institutions such as banks to intervene.P2P platform because of its own characteristics,in the borrower and lender borrowing process does not require third parties such as banks to join the risk prevention and control platform is mainly dependent on the personal credit rating system,but our country is not yet clear legal supervision measures,Platform for the borrower's credit information collection costs are high,the public credit system is not perfect.Owing to the asymmetric information,the borrower clearly understands the purpose of borrowing,willingness to repay,ability to repay,etc.,and the lender does not have all the information about the borrower.The moral hazard and the deportation of good money in this situation Are very common,resulting in frequent default fraud by borrowers.Lenders can not accurately determine the size of the risk of the loan,the platform can not be trusted,and ultimately may lead to inefficient P2 P market operation,over time the P2 P industry is bound to be severely hindered.By the end of 2016,the number of P2 P platform reached 2896,involving more than 18 billion yuan of funds.In view of this practical problem,this text has carried on the research to the fraudulent problem of the internet credit platform in internet finance with the combination classification algorithm in machine learning,and has made the following research results mainly:1.In order to effectively realize the identification and prediction of fraud in internet finance,based on the deep research on the distribution characteristics of fraud data in traditional financial credit institutions and the data traces of fraud clients under the background of big data,The machine learning model of online credit anti-fraud establishes the basic framework for the identification and detection of the fraudulent customers in the network credit platform P2 P.In order to make full use of the advantages of personal information widely distributed in the Internet in the background of big data,this paper designed and implemented the personal portrait model technology of fraudulent client based on big data in the approval strategy based on the traditional credit financial institution forms.In order to improve the accuracy of the classification model to identify and detect fraudulent customers in internet finance and maximize the coverage of fraudulent customers' anticipation,based on the AdaBoost algorithm based on combination of machine learning and two types The practical significance of false borrowing Proposed a comprehensive analytic algorithm for improving AdaBoost,which is mainly concerned with fraudulent customers.Based on AdaBoost efficient combination classification learning algorithm in machine learning,this dissertation designs and implements the identification and detection model of big data-based network credit fraud and applies it to specific and actual internet credit financial institutions to detect customer fraud In the relevant test.Through empirical analysis,this paper verifies the effectiveness of the technology studied in this paper in identifying and detecting fraudulent customers in online credit.