Research On Anti-fraud Technology Of Electronic Banking Channel Based On Semi-supervised Learning

In recent years,the Internet industry has developed rapidly.More and more Internet finance companies have emerged,and the models are emerging one after another.Generally divided into two camps: traditional financial institutions,such as direct banking,joint loan business.The second category is Internet finance companies,such as P2 P platforms,online lending platforms and consumer lending companies.The former is due to the upgrade tool but the risk control technology fails to synchronize,which leads to the risk,while the latter has a radical development,and the management system has a large number of loopholes.For financial institutions,identifying fraudulent transactions has an important impact on business development.Exploring a more effective and accurate anti-fraud model is one of the important development directions.However,due to the lack of tagged data,in many cases,supervised learning methods cannot be used for fraud detection.Fraud detection is a relatively complex task.Its development direction has not been uniformly determined in the academic world.Unsupervised methods such as clustering and outlier detection techniques alone cannot obtain satisfactory results.Another problem is the lack of sufficient business knowledge,so that the lack of interpretation and understanding of different customer behavior information,which also leads to poor results of fraud detection,making the fraud detection system unusable in the real environment.The research of this paper is mainly aimed at the research of commercial bank's electronic channel anti-fraud technology.Taking the semi-supervised learning as the starting point and combining the past bank anti-fraud experience,it establishes a model that can effectively solve the security problem of commercial banks' electronic channels.This paper focuses on literature research and project research,summarizes the improvement ideas of semi-supervised learning in various schools,and lays the groundwork for proposing the anti-fraud model of the bank electronic channel.Then,combining with the current bank anti-fraud practice,it integrates the anti-fraud framework and proposes how to combine the semi-supervised algorithms of various genres with the scheme.At the end of the study,combined with the actual transaction data,the semi-supervised learning is substituted into the anti-fraud model for verification.Based on semi-supervised learning,this paper proposes a variety of new fraud detection methods.This method takes into account the uncertainty of the transaction and injects several behavioral evidences of a transaction into the model.The behavior of the transaction is modeled by considering the trends of different principal and aggregate variables in different periods,and the behavior of newly arrived transaction events and historical transaction records is combined for behavior recognition.The final prediction is the combined result of using the proposed method combined with a large number of models.Finally,using the actual transaction data set of a commercial bank,the results of the new method are compared with the results of Logistic,Random Forest,CNN,etc.from the number of frauds found and the number of false warnings issued.The results show that compared with Logistic,Random Forest,CNN and other methods,the proposed method has higher accuracy and lower false positive rate,and the computational complexity of this method makes its response time longer.