Anti-fraud Analysis And Modeling Based On Social And Consumption Data

With the rapidly developed of Internet and the Internet big financial data,as well as the arrival of Internet + era,a large number of Internet financial net loan companies are appeared.Meanwhile the risk of fraud become an important issue in financial aspects of the Internet,how to effectively anticipate and identify potential user of fraud becomes the important target of the current Internet banking.By analysis the data on the social network and the consumption of a company,a more accurately predict and control model for fraud risk is proposed,to reduce the losses caused by fraud.The data are cleaned.Thereafter,to discover the relationship between consumer behavior and fraud,as well as the relation between social network and fraud,some descriptive statistics are explored,also the relationship between consumption and user fraud the relation between fraudulent users are discussed,then the appropriate variables are selected upon the results;The data are trained by the combination method of data mining machine learning algorithms and modern financial theory;The comparison and the evaluation are explored,and the optimal model is found upon the evaluation results by ROC curves.