Research On Credit Score Of Mobile Network Operator Customer

Credit management and control has become one of the important works of modern enterprise internal management.With Network Operators market scale expansion,Network Operators increasingly fierce competition in the market,followed by the customer's bad debt and "default" behavior also seems to be common occurrence.Therefore,the Customer Credit Risk analysis and evaluation of Network Operators are very necessary.A Customer Credit evaluation research of China Mobile Network Operator's Customer in a city,to effective risk control with the customers,improve the enterprise's ability to avoid credit risk,finally for our country Network Operators to carry out risk management to improve using for reference.Firstly the paper discusses the research background of the article,the research status at home and abroad,the research significance,and the research methods and thesis structure.Secondly,it describes the concept of credit,the 5C principle of customer credit evaluation theory and related content of credit evaluation.Then,it introduces the random forest algorithm for selecting indicators,the entropy principle and MDLP criteria,and the two models commonly used in credit scoring models,including logistic regression model and Fisher discriminant analysis model.Finally through considering customer defaults,customer profiles and a variety of factors such as consumer credit information and using SPSS software,the paper conducted empirical research.Due to too many indicators,this paper used R to establish a random forest to complete the selection of indicators.At the same time,there are continuous indicators and the indicators must be discretized.Obviously,the traditional discretization methods include equal-frequency and equidistant methods cannot meet the requirements.This paper used the attributes-entropy and MDLP criteria.This method is one of the best discretization methods at present.Finally using SPSS software,this paper carried out the Fisher Discriminant Analysis Model and the Logistic Regression model,the accuracy of the Logistic Regression model is over 95%.Because the Logistic regression has better stability and accuracy,we used the Logistic regression model designing the credit scoring.Through the analysis of test samples,the distinction degree achieved a satisfactory level.It shows that the model has very strong application value.On the basis of the set of quantitative criteria,the customer was divided into five types of credit rating.This paper used the real and effective data to put forward the index system and scoring model of credit rating of mobile users.So the work of this paper can help enterprises to form a reasonable credit evaluation system.Scorecard build will make enterprises can forecast customer potential default level,the implementation of customer service and reasonable order,improve the level of profit.Through customer segmentation,will help the enterprise marketing personnel according to the customer's different property distribution,accurately grasp the characteristics of customer group and personalized requirements,carry out reasonable marketing plan,and improve the operational efficiency of the enterprise.Enable enterprises to make the comprehensive quantitative judgment of Customer Credit Risk,according to credit rating for the client to take pertinent measures to guard against the customer's credit crisis.