Design And Implementation Of Customer Retention System Based On Data Mining Technology

With the rapid development in the industry of electronic communication, the competition among Chinese major operators is becoming more intense. There are three major operators in China (China Mobile, China Unicom and China Telecom), all of which have their own customers. But at the same time, they all face with the problem of rising customer loss rates, mainly because existing solutions don’t apply the theoretical basis of data mining, and existing systems mostly take a certain city as a tryout. Therefore, it lacks universality and expansibility, and the channels can’t be fully expanded.To solve the above problems, the decision tree method is used to establish the customer group model in the customer retention system. We design mutual-exclusion matrix rules as well as preferential control rules in order to improve customer retention rate. Specifically, the main work includes the following four aspects:Firstly, the overall architecture of the customer retention system is designed. The customer retention system is not an isolated system, which needs interactions with other systems constantly. The customer retention system acquires basic data from the data center of the city and then it can provide basic attribute information of customers for marketing. The customer retention system extracts data of customers from the tag library, and then pushes these data into the system. Meanwhile, the customer retention system can be entered in through the operation analysis system, which can provide a unified entrance.Secondly, the customer group model is designed. We use the decision tree of the data mining theory, and establish attributed nodes based on the ID3 algorithm. Therefore, customers are classified based on their basic properties, and the customer group model is constructed.Thirdly, the method of policy recommendation is designed. This thesis filters the recommended policies by using mutual-exclusion matrix rules. If the policies are exclusive or have already been recommended, they will not be recommended, otherwise vise versa. The policies that use the preferential control rules will effectively avoid getting excessive discount for individual and customer group, thus the vague boundary of repeating preferential control will disappear.Finally, the customer retention system is implemented. This thesis designs each functional module of the customer retention system. First, this thesis integrates the methods of customer group model and policy recommendation, and studies the influence of various channels on the customer retention system. Then, the customer retention system adopts the technology of web service during the interactions with other systems, and thus effectively improves its real-time and security. In addition, we complete the development of the customer retention system, and verify and test it.In conclusion, the thesis designs and implements the customer retention system, which uses the data mining technology to build the customer group model. We construct the customer group model, and design the method of policy recommendation, which filters the nearly recommended policies by using the mutual-exclusion matrix rules and preferential control rules. Ultimately, the efficiency of customer retention will be improved and we can retain customers to the greatest extent, thus the long-time stability of customers can be achieved.