Credit Evaluation Data Cleaning Algorithm Based On Social Network Data And Its Rapid Implementation

With the rapid development of Internet finance,as a useful supplement to traditional financial industry,it provides relatively effective financial services for lowand middle-income groups.However,due to the lack of credit evaluation methods,credit risk is still the main source of risk in the Internet financial industry.The construction of natural person credit model and credit evaluation are effective means to reduce credit risk.With the widespread use of online social tools,more and more human individual behavior has been faithfully recorded,forming a huge social network database.These data record real human activities and are part of the social mapping of real people,making it possible for social network data to be used to measure users' credit level.In this paper,we mainly carry out two aspects of research work.Firstly,in the research of social data as supplementary data of personal credit evaluation model,there are often anomalous nodes in the network which are not enough social footprint or can not represent ordinary real users.The existence of these nodes affects the ranking results of credit evaluation.Therefore,we build a data cleaning model for personal credit evaluation of social data based on user's degree distribution,activity and time interval of user's behavior.We rank the data sets before and after cleaning,and observe the impact of cleaning model on the results.Secondly,we test the whole cleaning model with use cases to find out the most time-consuming part of the whole cleaning process and accelerate it in parallel.In the first chapter of this paper,we introduce the research background and current situation of social data used in personal credit evaluation,and point out the existing problems and the work we need to do.The second chapter mainly introduces the related technology and methods used in this paper.The third chapter is mainly about the acquisition and description of the research data.In the fourth chapter,we elaborate the construction and principle of the whole cleaning model,and calculate and sort the data before and after cleaning with the personal credit evaluation model.The main work of Chapter 5 is to find out the most time-consuming steps in the whole cleaning model,and accelerate the calculation to improve the response speed of the cleaning model.In this study,the cleaning model based on user characteristic attributes and behavior attributes is the focus of this paper.The innovative points of this paper are the distribution of bonding degree,activity degree and cleaning conditions established by behavioral time interval.Finally,the cleaning model we got can clean the "real and false data" in personal credit evaluation.