Research On Consumer Finance Credit Rating Model And Application Based On Data Warehouse

With the advent of Internet+ era,consumer finance has gradually become an important part of Internet financial services.Traditional credit risk assessment models are less adaptable because consumer finance companies face middle and low-income,lack of credit-history customers.Big data technology is a major technology change in IT industry,in the era of data explosion,big data improves the risk identification ability of unsecured credit loan industry effectively.The data has become the core financial institutions,driven by innovative technologies such as mobile internet,social networking,and big data technology.When using Internet,a large number of information,customer personal characteristics data,e-commerce data,social data payment data etc can be centralized,integrated and processed,modeling to carry out credit rating,forecasting risk to assurance company sustainable management.Through the application of big data technology and cloud computing,it's possible to establish a multi-level and multidimensional credit risk assessment system.The thesis aims at the Internet consumer finance risk control as the research object,aiming at the urgent demand of the network consumer credit loan to the big data risk control and anti-fraud,puts forward the systematic big data financial empirical research scheme for the network loan,and establishes the network loan credit risk evaluation model on this basis,enriches and consummates the credit risk management theory and the method.Aiming at the problem of general shortage and sparse sample attribute of consumer finance company,this thesis proposes a training method of small sample data of consumer finance based on generative anti-network(GAN),and expands sample quantity by producing customer samples with same characteristics as real samples through GAN to solve the problem of insufficient sample data in the early stage of consumer finance enterprises.Using K-means Clustering algorithm,this thesis trains the customer data of an Internet consumer finance enterprise,obtains the customer subdivision model,and through the subdivision of the customer classification,understands the attribute characteristics of different classification customers,and more targeted implementation of the risk control strategy.A number of algorithms,such as logistic regression and random forest are used to train the credit limit of the customers,and finally obtain the probability model of customer's rating.The model can be used to predict the user's loan quota according to the risk attribute,personal attribute and other dimensions,which can help to carry out the main business such as pre-loan approval and credit adjustment.Based on the above model,a kind of data warehouse and its private cloud architecture supposing real-time high concurrency are designed and implemented,which can integrate various distributed data parallel processing framework,and proposes the corresponding task scheduling for the practical characteristics and application scenarios of sparse in consumer financial data.It improves the efficiency of data read-write and analysis with good openness and expansibility.