Big Credit Data Risk Control Based On Federated Learning:Research And Application

In the field of pan finance,risk control is one of the most important links within any credit business.With the rapid development of computer technology,it has become a consensus in the industry to use big data for risk control modeling.At present,credit big data risk control technologies mostly rely on the exchange and integration of user privacy data between different enterprises.However,with the gradual strengthening of domestic and foreign data supervision and privacy protection,this data fusion method that seriously leaks user privacy will no longer allowed.Aiming at the data island problem,we explore the establishment of a federated learning system in the credit risk control scenario,and based on this,we design a risk control model to improve the risk control ability of the enterprise.Firstly,we analyzes the current situation of big data in the credit field,and proposes solutions for unified data access formats,preprocessing,and federated modeling in view of the problems of data heterogeneity,data privacy,and security;then,The feasibility of federal learning in the field of risk control was explored,and a federal learning system suitable for credit risk control scenarios was constructed.Finally,based on this,an Embeddding Sequence Recurrent Neutral Network(ESRNN)based on risk sequence embedded representation.The model is based on a deep understanding of the domain data.By modeling the user risk sequence mode,it can effectively overcome the shortcomings of the previous risk control model for the insufficient modeling of serial data,thereby improving the risk control effect.The paper uses two methods to evaluate the effectiveness of typical risk control approaches on two real credit scenario datasets.The results show that compared with previous methods,ESRNN-based credit risk control models are used in AUC,KS and ACC have achieved the best performance on the evaluation indicators,thus verifying the validity of the research in this paper.In the future,we will focus on the mining of user multimedia data in the federal scenario and the interpretation of risk control models to help credit companies achieve better risk control effects under the premise of data compliance.