Deep Learning Technology For Financial Industry Monitoring And Auditing Scenarios

With economic development and social progress,bank cards have played an increasingly important role as an important means of payment and capital carrier[1][2].At the same time,the related business of bank card is also the key part of audit and monitoring in the banking industry.As the main participant of bank card business,the issuer and the card swiping merchant are also in the regulatory system of China UnionPay.In this paper,according to the two business backgrounds of the bank card issuing behavior and the merchant's credit card transaction auditing and monitoring,two automatic machine learning auditing frameworks are proposed based on the fact that the current commonly used manual auditing is inefficient in estimating and judging.For the bank card audit scenario,we propose an automatic machine learning framework for bank card audit.It includes four parts: the acquisition module of bank card audit data,the element measurement module based on deep learning,the dimension audit module based on machine vision algorithm and the overall audit induction module.In particular,we transform the card detection task into the problem of target location in deep learning detection and traditional machine vision.At the same time,in order to improve the efficiency of model training,we transform the training of automation framework into a super parameter black box optimization task,and use it to make the model quickly find a better training.For the transaction monitoring and auditing scenario,we propose a machine learning problem that formalizes the detection of illegal merchants into a graph convolution model.Innovatively,the merchant relationship is represented in the form of graph structure data,and the implicit relationship of node features is extracted by graph convolution network,which shows better than the existing method.