Research On Customer Churn Prediction And Retention In Telecommunication Based On MMOI Method

Customer churn is not a simple problem whose object is to retain customers, but a complex system involving many factors, such as operators, customers, government and technology, which influence each other. In customer churn prediction, there exist many characters, such as data sources being numerous, data attributes relationship being complex, the number of categories being asymmetric, and so on. In customer churn retention in telecom, there exist various kinds of effects influencing customers churn and customers retention, meanwhile, the telecommunication enterprises must consider the conditions including the limitation of the internal resourses and the response of the external rivals. However, the existing researches on customer churn still lack of a set of scientific, system theory and method; the existing methods for customer churn prediction based on single models cannot completely meet application needs; the existing researches on customer churn retention in telecom based on a summary of strategy and qualitative analysis cannot give much guidance for telecommunication enterprises when making scientific retention stragies. Therefore, it is of important theoretical and practical contribution to explore and study a new theory frame and method system for analyzing customer churn in telecom and to construct a kind of efficient predition models and a kind of scientific retention models for customer churn in telecom. Based on the substantive characteristics of customer churn in telecom, the theory framework and method system for analyzing customer churn in telecom are studied. Then, around the targets of lifting the ability of predicting customer churn and optimizing strategies of customer retention in telecom, a series of researches are launched for customer churn prediction and retention in telecom.First of all, customer churn management in telecom is a complicated problem, and there is little mature theory to guide management practice. Inspired by the existing ideas to solve the problems of complex systems (integrated thoughts, the thoughts of model integration, and the thought of system dynamics), the theory framework and method system are presented to analyze customer churn in telecom based on Multiple Models Optimized Integration (MMOI). The framework consists of customer churn prediction analysis module and customer churn retation analysis module in telecom, the former mainly increase the prediction performance of the system through paralleling the multiple prediction sub-models by optimization, and the latter solves effectively the complex problem of customer churn retention in telecom mainly by separating the complex problem into pieces and modeling, respectively, then integrating the models in series. Research results show that the theory framework and method system analyzing customer churn in telecom based on MMOI proposed in this paper can guide the customer churn problem effectively solved from theoretical height.Secondly, the problem of customer churn prediction has the features of data sources being numerous, the relationship among data attributes being complicated, and the number of the categories being asymmetrical, however, the existing researches based on single prediction model cannot satisfy accurate requirements when predicting the customer churn in telecom. Therefore, guided by the theory framework and method system analyzing customer churn in telecom based on MMOI proposed in the second chapter, combining different integration goals (based on the prediction precision and forecast earnings), different integration ways (linear integration, nonlinear integration, and dynamic integration), different intelligent optimization techniques (Artificial Bee Colony Algorithm, Artificial Fish Swarm Algorithm, Artificial Ant Colony Algorithm, Particle Swarm Optimization Algorithm, and Genetic Algorithm, etc.), we construct a series of customer churn prediction models in telecom based on MMOI. Researches show that not only predict performance of the customer churn prediction model in telecom based on MMOI excels common single models, but also the prediction results is more stable.Thirdly, the problem of customer churn retention has the features of internal relations among the effects being complicated, intend of customer retention changing dynamically, and the limitation on internal and external conditions of the telecommunication enterprises being complicated, however, the existing researches are mostly based on strategies summary and qualitative analysis and cannot scientifically guide telecommunication enterprise to retain the customers effectively. Therefore, guided by the theory framework and method system analyzing customer churn in telecom based on MMOI proposed in the second chapter, based on the actual situation of the limited resourses used to retain customers in telecommunication enterprises, we construct customer churn retention analysis model in telecom based on limited budgets. This model analyzes three effects (retention incentive effect, nature churn effect and word-of-mouth spread effect) affecting customer retention and the coefficients of the effects in detail from the angle of system dynamics, derives the formula calculating customer retention cycle and customer retention values according to customer retention rate, and then establishes the model of single customer retention and the model of one-on-one customer retention. On account of the actual situation of that retention strategy of telecommunication enterprises may cause competitors to counterattack, we construct a customer churn retention analysis model in telecom on basis of rival counterattack effect. The model analyzes four effects (retention incentive effect, nature churn effect, word-of-mouth spread effect, and rival counterattack effect) affecting customer retention and the coefficients of the effects in detail from the angle of system dynamics, derives the formula calculating customer retention cycle according to customer retention rate, and then establishes the customer retention values models and the customer retention strategies models under three condtions (having competition effect, having no competition counterattack effect, and competitive counterattack effect being uncertain). Research results show that the customer churn retention model can guide scientifically the telecommunication enterprises to establish customer retention strategies.Finally, from the angle of theory contribution and management practice, we analyze and evaluate the theory framework and method system of customer churn analysis in telecom based on MMOI, the customer churn prediction models and the customer churn retention models in telecom based on MMOI, their characters, and so on.