A Study Of Anti-fraud System Of Credit Card Transaction Based On Deep Learning Technology

With the popularization and development of credit card in our country, credit card payment has become one of the most important ways of payment. However, the huge market is the criminals' coveted goal. More and more counterfeit cards fraud cases and constantly updated ways of crime have made great serious financial losses, disrupted the normal financial orders and restricted the long-term healthy development of the credit card industry. Therefore, how to effectively prevent the risk of transaction fraud has become the concern point of domestic banks.The credit fraud includes application fraud and transaction fraud. With the application of credit card scoring data mining, the application fraud has been effectively curbed. The transaction fraud is more complicated and fickle and is increasing seriously in recent years. Data mining methods also have a broad application prospects in controlling the credit card transaction fraud. The neural network technology is applied to trade fraud detection research and achieved certain effect, however, diffusion gradient neural network can only defect the shallow network training, and its application is also limited.Recently, deep learning has made great development as a new artificial intelligence method, which solves the shallow defects of neural network, and can build the multilayer nonlinear relationship and makes great progress in image recognition field. Based on the domestic and foreign existing transaction fraud detection models, this paper used the deep learning technology in transaction fraud detection, constructed a transaction fraud detection prototype system based on DBN, and finally tested the transaction fraud scoring model based on DBN by using the data of actual credit card transaction, and confirmed its feasibility and validity.Firstly, the paper introduced credit card, fraud risk concept and its classification, and then elaborated the detecting principle of the credit card transaction fraud scoring model, studied the domestic and foreign popular transactions anti-fraud models, and described the deep learning technology as well as its application in DBN model. Finally, based on several steps in real transaction data including variable derivative, principal component analysis and data pre processing, the paper tested the effects of the model in detecting the transaction fraud, which provided the banks with the theoretical and practical reference in using the deep learning technology to detecting the credit card transaction fraud.