摘要：Churn is a defection from a service provider. This study explores the role of social network in customers' churn behavior .As the saying goes; customer is the god in service-based industries. Since the cost of customer acquisition is much greater than the cost of customer retention, customer churn prediction is more and more important. So far, research on churn prediction has involved user profile factors and social factors, but the analysis lacks interpretability and predictive horizon. In this paper, for a better understanding of churn behavior in MMORPG, we use longitudinal statistical models to examine whether the churn behavior in one person is associated with the churn of his friends on a complete data set of an entire society. Discernible clusters of active persons and inactive persons are visible in the network, and the relationship between people's churn behavior extends up to three separations. Besides, we provide quantitative evidence for the strong ties hypothesis. The higher the number of common friends two persons have, the influence between them is more significant. At last, we use mixed effects cox models to predict churn in the future. Our study can provide a long forecast horizon, it is an obvious advantage to prevent the churn of a person.
关键词：Social network; longitudinal analysis; three separation; homophily; influence
