ITDT: An Iterative Decision Tree-based Approach for Telecom Customer Classification

With the fast development of mobile technologies, mobile marketing has become an import task for telecom operators. As a result, customer classification has attracted the attention of many scholars and companies. Decision tree is a computational intelligence technique having been widely used in the field of machine learning and data mining. To solve the customer classification problem, we propose an iterative decision tree-based (ITDT) approach. Through the communication data provided by one of the largest telecom operators in China, we aim to identify student customers from non-student ones. We first build the customers' social network through their phone call and SMS records. Then we make statistical analysis on the network, which helps us to determine the network-based features used for customer classification. Last, we propose an iterative decision tree-based algorithm to identify the student customers. We conduct experiments under different algorithm parameters and the results show the effectiveness of our algorithm.

I. Introduction

In the era of intelligence and big data, the fast development of telecom infrastructure and popularity of mobile devices have witnessed the prosperity of mobile social networks. The broad market space behind the mobile users has attracted the attention of many enterprises and organizations, resulting in the proliferation of a new research area, i.e., mobile marketing. For example, by mobile marketing, telecom operators may identify potential customers to promote value-added services [1]. Other examples include maximizing the adoption of information or new product through large-scale mobile networks [2]–[4], attracting the maximum number of mobile users to visit a location [5], etc.