Predicting customer churn by integrating the effect of the customer contact network

Customer retention is one of the most important tasks for many enterprises. In order to retain customers, customer churn prediction is necessary. This paper investigates the effects of network attributes on the accuracy of predicting customer churn. The contributions include the following: (1) a relatively complete set of network attributes are provided and incorporated into prediction models that are built using machine learning algorithms; (2) the effects of network attributes on prediction accuracy are measured while including the determinants proven to be effective by other researchers, and several models are constructed and compared to assess model effectiveness. This study uses a customer data set of a Chinese mobile telecommunication. The results show that the network attributes can greatly improve the prediction accuracy.