Churn Prediction by Finding Most Influential nodes in Social Network

The rapid growth of telecommunication industry has also elevated the growth of social network resulting in more number of people connected to each other. Social network is a platform where people share their views with each other, which leads to influencing one another for buying a product, stop using any particular service etc. A user who stops using a service is a churn user and the process to stop using a particular service is called churning. The existing work on churn prediction does not consider the social aspect of the customers. In this paper we propose to use social network information of the customers along with their call log details to predict the churn users. The set of most influential nodes is predicted by using the social network data, these predicted churn users may be the one who may influence the others to stop using a service. A proper retention policy if applied to these customers, can prevent the customers from churning a service. The experimentation was done by using Pokec social network data and generating synthetic call log details of these social network users. It was observed that the accuracy of churn prediction is improved when combining social network and call log information of the users for churn prediction.