Research And Implementation Of Credit Transaction Risk Assessment System Based On Neural Network

With the rapid development and deep integration of Internet technology and finance,a variety of online credit products continue to bring forth new ideas and have entered People’s Daily life.Different from the traditional bank loan business,online credit has gradually become the most commonly used loan channel by virtue of its advantages of convenient application process,flexible loan amount,and can be handled anytime and anywhere.When developing network credit,how to predict in advance whether users will default and how to effectively reduce the losses caused by bad debts are the key problems facing all credit institutions.Therefore,it is of great significance for the long-term development of Internet credit institutions to establish a credit transaction risk assessment model that can effectively predict user default risk.Aiming at the above requirements,this paper proposes a Jumping Splicing fully connected neural Network（JSNet）credit transaction risk assessment model,and applies the model to the developed credit transaction system.The main work is as follows:Aiming at the unbalanced credit data set with large difference in the proportion of positive and negative samples,a t-test-weighted oversampling method is proposed,and statistical indexes are introduced to comprehensively consider the data distribution pattern and the significance of the difference between positive and negative samples,so as to reasonably generate a few samples to achieve the purpose of sample equilibrium.Then the processed data are input into the credit transaction risk assessment model based on JSNet neural network proposed in this paper to predict user default risk.The model introduces residual network idea and adopts iterative jump structure,which is beneficial to model fitting more information and avoiding overfitting.Compared with the three mainstream models on the open Lending Club data set,the experimental results show that the model proposed in this paper has improved in all indicators.According to the demand,this paper develops a credit transaction risk assessment system based on neural network.Based on the investigation of the current credit transaction platform,it firstly analyzes the demand of the system from the perspective of users,functionality and non-functionality,and then makes an overall design of the system by using the system framework structure diagram and other description methods.Next,the system is designed in outline and detailed design of each specific functional module and database design.In terms of system technical architecture,the backend uses The MVT mode of Django framework in Python to develop,and the front-end uses vue.js framework to construct the system page,and finally realizes the credit transaction risk assessment system.In addition,the user portrait page is added,and users are grouped by k-means clustering algorithm to visualize the attribute characteristics of loan users.Through the function test and performance test of the credit transaction risk assessment system,the results show that all the functions of the system have achieved the expected effect,and can meet the requirements in the aspects of security,ease of use,concurrency and other performance.