Telecom Customer Segmentation Model Design:a Big Data Perspective

Telecom customer segmentation model design in a big data perspective has significant importance in both research and commercial fields. Research on big data computation has become a key area in Project973of China and behavioral analysis on media data becomes a focus in Project863.To this end, three main contributions are made in this paper. First, this paper research deeply into clustering techniques according to business characteristics. Then, clustering algorithms are introduced and designed for better efficiency in real telecom customer segmentation applications. Finally, a Prototypical Solution for telecom customer segmentation in a big data perspective is developed.Specifically, the main work are listed below in this paper:Firstly, on the basis of referring literature and analyzing customer segmentation problems, algorithms based on k-means clustering is proposed in the telecommunications customer segmentation, and thus a whole framework for the entire customer segmentation application is suggested. In this paper, customer segmentation is desigined both on the macro view and micro view called subgroup analysis. This chapter dives deeply in customer electricity data, focusing on the analysis of data sources, data preprocessing, data integration. In addition, value and social attributes are introduced to build the key elements of customer segmentation program.Secondly, a prototype system of big data telecom customer segmentation system is designed and implemented. The overall system follows a hierarchical and modular design criteria, building three-tier architecture, including data layer, model layer and application layer.In the overall functional design, three main modules are designed including a basic customer information management, cluster analysis, social network analysis. For system implementation, the rapid prototyping programming language-Python is used, and a Python-based web framework Django is implemented within the B/S system architecture. Whats’more, NoSQL data modeling techniques are introduced and a mixture of several mainstream NoSQL databases are implemented in the system.The main contribution of this paper includes the following aspects:1. using clustering techniques of data mining theory, this paper digs into customers through basic information and consumer behavior information, and outputs clustering rules, and after customer segmentation it uses precise marketing positioning to develop new customers and retain old customers;2. this paper designed and implemented a telecom customer segmentation system with a good prospect;3. NoSQL data model is introduced according to the mass volume of data, focusing on solving storage problems on saving user behavior information.