The Research Of Personal Credit Scoring Model Based On Logistic Regression

Consumer credit is that banks or other financial institutions provide consumers with personal loans by using credit,mortgage and other means.Due to various factors such as the level of macro economic development,consumption concept and other factors,China's consumer credit business started late,at the end of 1990s,before entering the fast track of consumer credit development.Its scale has been continually expanded.Central bank data shows that the size of China's consumer credit reached 19 trillion RMB in 2015,an increase of 23.3%.However,compared with developed countries,China's consumer credit development level is still very low,the proportion of consumer credit in the total consumption is still less than 1%,the proportion of western developed countries generally 20%-40%.But from the point of view of development,the development trend and potential of consumer credit will be huge,especially with the rapid development of Internet Finance,customers can quickly apply for loans through a simple audit.This pattern of consumption in line with the current people's living habits,has brought great convenience to people,it can be predicted that the development potential of consumer credit in China will be huge.In order to adapt to the rapid development of consumer credit,effective risk management is an essential part in the control of risk.It is important for the safety of credit funds of the company and the final profit by the development of credit scoring model to strength risk management of personal consumption credit and reduce the risk of consumer credit.This paper firstly reviews the basic concepts and research status of credit scoring model,and makes a comparative analysis between the two aspects of the application conditions and the accuracy of the modeling methods of credit scoring models.Secondly,taking an online loan product of an internet financial company as an example,we develop a special scoring model for the product.The development process of the model mainly includes data collection,pretreatment,the creation of scoring model,model test and monitoring.In the process of model training,this paper tries to use the method of cluster analysis to speed up the training process of the model.Finally,the effectiveness of the credit scoring model is analyzed by measuring the effectiveness of the model,the distinction between the degree and stability.After the research and development of credit scoring model in this paper,we have following conclusions.Firstly,credit scoring model based on logistic regression method in line with our current consumer credit industry and meet the current data requirements,and has the advantages of easy implementation,good explanation etc.Therefore,this is the main modeling method to establish credit scoring model in china.Secondly,in the process of model training,through the clustering analysis method to eliminate the statistical significance of repeated variables,we can effectively speed up the process of model training.Finally,according to the special customer group,combined with the business needs,it is necessary to establish a special credit scoring model for specific products.It should be noted that,in the development of a product specific scoring model to ensure that there is sufficient amount of customers and data.Because the amount of data is not enough,in the definition of the target variable will be biased.In addition,the credit scoring model can be improved from many aspects.The first one is the diversification of the data source.With the development of big data,the data is not limited to the credit data,but also relates to the Internet,mobile Internet,social network data,third non financial behavior data,helps lenders make more reasonable and comprehensive credit score,from many aspects of the customer's repayment ability.Therefore,with the improvement of theory and method,the credit scoring based on big data analysis is the research focus of the future model development.The second one is the improvement of modeling method.With the development of China's consumer credit industry,we can also try to establish a credit scoring model with the method of artificial intelligence.In addition,the credit scoring model can also quantify the impact of decisions,from two aspects of customer credit performance and credit decision-making to predict the future performance of customers.This is also the direction of the credit scoring model in the future.