Research & Practice Of A Method On Customer Credit Evaluation In Mobile Communication

Along with the diversified development of mobile communication business, the problem of personal credit is getting more and more attention of the operators. At present most of the mobile operators have few records of customer's credit, as long as customer's telephone charge is overdue, they will be out of service. This measure may bring about the increasement of customer complaints but decreasement of the capacity of service level. Worst of all, it may causes a serious problem of customer churn; also, it will hinder the diversified development of mobile industry and the raise of telecoms operator's competition ability. So the operators need to find an accurate way to evaluate the customer's credit according to their basic attributes and behavior. Then the carrier can offer different services on the basis of the users'credit.The principal research work and innovations are brought forward, including:1. This paper investigated and studied the reason and background that caused the fee loss in China communication market, also this paper discussed the situation of the customer credibility management in mobile.2. According to the existing business rules and business scene in mobile communications, the article analyzed the factors which influence the reputation of mobile customers.3. As there are a great number of customers and the information of them may be changed relatively frequently, compared with other credit evaluation method, this paper put forward a model based on the algorithm of Information Gain and Bayesian theory to analysis and evaluate customer credit. The model need to preprocess, the untreated and comprehensive date from the database of telecom operators such as clean up, integrate, transform and other pretreatment. Then it will calculate the information entropy of each attribute to determine which attribute can bring much information for classification system using the method of information gain. 4. In order to forecast the level of customer's prestige, this paper put forward the method of Naive Bayes Classifier.Finally, this article has done some empirical analysis to the model with the real date of mobile customer, Beijing City; and list the process of system implementation and test results. The results indicated that the prediction deviation ratio of this model is low, and these results proved the feasibility and effectiveness of this study.