Research On Distributed Credit Risk Control User Portraits

With the rapid development of the Internet,people's daily work and living habits have undergone qualitative changes.The development and application of Internet technology has penetrated into all walks of life in a lightning-fast manner,especially in the financial industry.The traditional financial industry is undergoing a severe test of unbalanced asymmetry in big data,the Internet,and user experience.Big data risk control technology has become the focus of a big data application that most people in the industry are concerned about.The financial risk control model is an inevitable trend for the healthy growth of internet finance in the sun.At present,China's Internet financial credit system is not yet perfect.The credit system and relevant laws and regulations contain certain defects.The Internet extracts personal static data and behavioral data for analysis and modeling,that is,credit risk control user portrait studies are risk control.The main content of the model.In the current aspect of Internet finance's risk control security,this paper presents a study of distributed credit risk control user portraits.This article first uses the user's credit data in recent years to model,user information is tagged,and then converted into multi-dimensional numerical data,in order to facilitate the computer to identify the processing.Secondly,with the explosive growth of data volume,classical data mining is far from meeting people's needs.K-Means is a traditional clustering algorithm and is widely used,but K-Means algorithm randomly selects the initial K.One center point is easy to make the clustering effect unstable.In addition,the K-means clustering algorithm is sensitive to noise points,and some points are scattered around.This often leads to a local optimum rather than a global optimum.Therefore,this paper proposes a K-Means algorithm based on kernel density,and at the same time,combines the MapReduce distributed architecture to perform distributed computing on the improved K-Means algorithm.The improved algorithm can effectively solve the problems of K initial center selection difficulties and noise point sensitivity.Thirdly,the analysis of the experimental results shows that the clustering effect of distributed K-Means based on kernel density estimation is significantly better than that of classical K-Means and distributed K-Means both in the number of iterations and the accuracy of the results.The class effect is much better.Finally,through this analysis and comparison,the results obtained by clustering are reduced by the user ID number,and the user's portrait is depicted in the form of a tag.As a result,the user is divided into different group categories in order to make credit risk control of the Internet finance well.