The Research of Customer Loyalty Improvement in Telecom Industry Based on NPS Data Mining

作     者：Lili Tong Yiting Wang Fan Wen Xiaowen Li 

作者机构：Beijing University of Posts and Telecommunications Beijing 100876 China School of Economics and Management Beijing University of Posts and Telecommunications Beijing 100876 China School of Computer Science Beijing University of Posts and Telecommunications Beijing 100876 China China Mobile Communications Co. Ltd. Enterprise Customer Branch Beijing 100032 China 

出 版 物：《China Communications》 (中国通信（英文版）)

年 卷 期：2017年第14卷第11期

页      面：260-268页

核心收录：

学科分类：0810[工学-信息与通信工程] 12[管理学] 0202[经济学-应用经济学] 02[经济学] 1201[管理学-管理科学与工程(可授管理学、工学学位)] 0808[工学-电气工程] 020205[经济学-产业经济学] 0809[工学-电子科学与技术（可授工学、理学学位）] 0839[工学-网络空间安全] 0812[工学-计算机科学与技术（可授工学、理学学位）] 

基　　金：Supported by Humanities and Social Sciences Foundation of Ministry of Education in China (Project No. 16YJA630063) 

主　　题：net promoter NPS customer loyalty data mining 

摘      要：In recent years, the telecommunications have used the concept of NPS(Net Promoter Score) for customer relationship management, but there is neither definite theory research nor instructive instance research. However, this paper summarizes an approach with instance case analysis to improve customer loyalty via NPS data mining, which has extensive and practical significance for tele-companies. First, this paper finds some driven forces of customer loyalty, which are relative to customer consumption such as the call duration, the usage of data, ARPU, etc., by using some innovative reasoning-analysis based on IG(Information Gain) and xg-boost decision-making tree model, so the tele-companies can predict the role of individual customer and form daily monitoring on big data, which will save a lot of NPS survey cost. Second, this paper summarizes how customer group feature impacts the relationship between NPS and financial performance. Taking ARPU value as the performance goals, we divide the sample customers into 6 groups and summarize their characteristics based on k-means clustering, and give targeted suggestion of each group.