الفهرس Only 14 pages are availabe for public view
 |  | from | 107 |  |  |
| | | from | 107 | | |
Abstract
Mobile cellular subscriptions are expected to reach more than 7 billion worldwide, by the end of 2015 [1]. As the mobile market gets saturated, it becomes harder for telecom providers to acquire new customers, and makes it essential for them to retain their customers. Due to the high competition between different telecom providers and the ability of customers to easily move from one provider to another, all telecom service providers suffer from customer churn (i.e. the loss of customers to other service providers due to un-satisfactory services). As a result, churn prediction has become one of the main telecom challenges. The primary goal of churn prediction is to predict a list of potential churners, so that telecom providers can start targeting them by retention campaigns. Typical churn prediction is done by analyzing customers’ data for low customer usage or bad service quality. A new promising trend based on social network analysis (SNA) is to model churn as a social behavior, where churning customers affect their social circles. In this model, customers and their call interactions are viewed as a social network, where customer churn propagates from one user to another according to the strength of their relationship (i.e. social tie strength). In this thesis we model churn as a dyadic social behavior, where customer churn propagates in the telecom network over strong social ties. We propose a novel method for measuring social tie strength between telecom customers. We then, incorporate strong social ties in an influence propagation model, and apply a machine-learning based prediction model that combines both churn social influence and other traditional churn factors