【作者】 陈金波；

【导师】 何建敏；

【作者基本信息】 东南大学 ， 管理科学与工程， 2006， 博士

【摘要】 面对电信市场竞争的加剧和信息技术的发展,电信企业必须建立以“客户为中心”的管理模式。因此,利用数据挖掘技术对海量的电信企业客户数据进行挖掘分析,从中发现各种潜在的、有价值的、规律性知识,是当前电信企业提升CRM水平的重要方面,极具理论意义和应用价值。本文运用理论分析与实证研究相结合的方法,针对数据挖掘在电信CRM中的若干个具体应用问题进行研究。主要内容如下:1.详细地分析了电信企业IT系统现状,建立面向客户主题的电信企业数据仓库体系结构,对电信企业数据仓库主题分析进行了研究,设计了相应的数据模型:物理模型和逻辑模型,并对电信企业数据仓库的实现方式进行了分析论述。2.系统地介绍了CRM理论,设计了以客户为中心、闭环的四层电信CRM体系结构;对电信客户管理进行系统地研究,以电信客户生命周期管理理论为框架,建立了基于数据挖掘的电信客户生命周期管理模型。3.依据CLV理论,建立了基于当前价值、增量价值和存量价值的电信客户价值模型;并以此为理论依据,设计了电信客户价值评价指标体系;结合AHP法,提出了电信CLV的计算方法,并对某电信企业客户进行了实证分析。4.建立了遗传算法优化的改进K-means(GLKM)聚类模型,研究了有指导的聚类模型评价方法,并进行了仿真验证;最后利用某电信公司客户数据进行了实证分析,并对客户群进行特征刻画。5.基于代价敏感学习理论,分别利用Under-sampling和AdaCost算法来构建代价最小化的电信客户流失预测模型,并通过总代价比较和模型收益性分析来表明代价最小化的模型具有更高的应用价值。本文的研究工作为电信企业应用数据挖掘技术分析客户行为和提升CRM水平可提供有益参考,在理论研究和工程实践上具有重要意义。更多还原

【Abstract】 Facing the fierce competition in telecommunication market and development of information technology, telecom companies must establish a management pattern with“centered on customers”. Therefore, it’s a significant aspect for telecom company to improve CRM level using data mining technology to mine and analyze a great deal of telecom companies’customer data and discover various of potential, valuable and regularity knowledge. It has theoretical meaning and applied value. Aimed at some application problems of data mining in telecom CRM, the thesis takes the study with the methods of theoretical analysis and empirical research. The object matter is as follows:1. This thesis analyzes the actuality of telecom companies’IT system in detail and establishes a telecom data warehouse. Then, the analysis subject, data models (physical model and logical model) and the implementation method of data warehouse is researched.2. The thesis studies the CRM theory by the numbers; designes the closed-loop and four-layer architecture of telecom CRM. Telecom customer management is researched by the numbers, and a telecom customer lifecycle management model based on data mining is brought forward under the phase of telecom customer lifecycle management theory.3. The three dimensions telecom customer value model based on current value, incremental value and perserved value is brought forward according to CLV theory. Based on this theory, telecom customer value evaluation index system and computing method combined with AHP method are put forward, and empirical research on the telecom company PHS customers is taken.4. A modified K-means clustering model optimized by genetic algorithms is established. In the data process and in results evaluation this thesis researches supervised clustering model evaluation method. Finally, this thesis takes an empirical analysis with the data of public customers in a telecom company and describes the character of clusters of customers.5. According to the cost-sensitive learning theory, this thesis makes use of under-sampling and AdaCost to build telecom customer churn prediction model with minimum-cost, and demonstrates that minimum-cost prediction model is better than traditional model through total cost and model benefit analysis.The thesis provides beneficial reference to telecom companies in analyzing customer behavior and advancing CRM level using data mining technology, and shows important meaning in theoretical research and engineering practice.更多还原