Study On The Active Service-Oriented Customer Churn Prediction Model

With the commodity market transition from a seller to the buyer's market and customer-centric system of scientific theory formation and improvement, the customer has become the most important and scarce resources in modern manufacturing enterprises. Customer determines the enterprise's fate and future in large part. While enterprises attach importance to our clients, they have increasing awareness of the importance of keeping customers, which increased customer churn management. Customer churn prediction will be the core content in customer churn management for active service.According to the active service and future-oriented service of Service-mining theory, and gradient and known ability of enterprise-class customer behavior, this paper combines improved qualitative simulation algorithm and improved genetic algorithm with artificial intelligent technology, designs customer churn prediction agent and adaptive modification agent for active service. So, it conceives the Active Service-Oriented Customer Churn Prediction Model, and develops the prototype system to stimulate, which tests the validity of the model. On the one hand, This model implements the intelligent and automatic churn management system, reflects the idea of mining services, provides the churn probability at each stage for each kind of customer, and provide some qualitative reference for how to make decision to get the best effect in enterprise-class customer churn management; on the other hand, it can make a positive role in the development of active service-oriented customer churn management system.This paper focuses on active service for customer churn prediction and adaptive modification. In the design of customer churn prediction agent, this paper contrives the customer churn factor system using causality graph and put forward the several concept about decision variables, state variables, constraint of sequential causality. And it replaces random behavior extraction of the one-off reasoning of traditional QSIM and avoids the combination explosion of algorithm. According to the analysis of experimental stimulation, it indicates that the predication agent can explain and predict the evolution of customer churn and provides reference to make decision and implement.In the design of adaptive modification agent, this paper analyzes and researches on the traditional genetic algorithm, and designs the modified immune mechanism, dual coding system and improved genetic operators. Therefore, it eliminates the system error of operating results and indicates the adaptability, generality, flexibility and diversity. After the appropriate experimental simulation, it shows that setting different parameters of the adaptive modification algorithm parameters and solving model parameters according to the characteristics of different industries in different environments can effectively modify the system error of different customer churn predication.