【作者】 齐先锋；

【导师】 喻金平；

【作者基本信息】 江西理工大学 ， 计算机应用技术， 2007， 硕士

【摘要】 数据挖掘是较先进的数据分析方法,可以对客户数据进行深入地分析。“以客户为中心”的先进的经营管理理念充分体现在客户细分上。客户细分最大程度地满足客户需求,使企业利润达到极大化。将数据挖掘技术应用于客户细分,能够帮助企业实现既定目标。本文针对电信行业的客户细分,主要做了以下工作:首先,介绍了国内外有关电信行业客户细分常用的方法,比较分析了各数据挖掘算法的特点。结合电信行业的数据海量的特点,重点分析了经典K-means算法和k-medoids算法的设计思想,确定了以优化初值、固定细分标准、提高执行效率作为算法改进的关键点,提出了适合于电信行业客户细分的tt-k-means(二次均值)算法和tt-k-medoids(二次中心点)算法,并完成算法的详细设计。其次,采集赣州电信公司的样本数据,结合部分经验值,对系统进行了验证,并对细分结果进行了比较分析。实验结果表明,tt-k-means和tt-k-medoids细分结果更准确,运行效率更高。最后,介绍了数据挖掘的整个流程和数据仓库的架构模式,在该系统架构之内,设计实现了客户细分系统。该系统以tt-k-means和tt-k-medoids为核心,由系统管理模块、预处理模块、聚类分析模块、知识表达模块构成。开发的系统具备试运行条件和应用推广价值。它的成功开发和应用,为今后实现其它类别的客户细分系统做了有益的探索。更多还原

【Abstract】 The advanced management idea of“Take the customer as the center”manifests fully in the customer segmentation. Customer segmentation may meet customer need at whole hog, enable the enterprise to achieve the maximal profit. Data Mining is an advanced effective data analysis method. Its Application to customer segmentation will help the enterprises to gain their ends on the best way.As for customer segmentation in telecommunication enterprises, in this article I have mainly done the following work:Firstly, It introduced commonly used methods of customer segmentation in the domestic and foreign telecommunication enterprises, various characteristics of data mining algorithm by comparative analysis.With emphasis I analyzed the design concept of the classical K-means algorithm and the k-medoids algorithm, determined optimizing the initial value and the fixed segmentation standard and efficiency of carrying out to take the algorithm improvement as the key point, proposed tt-k-means and tt-k-mediods algorithms, and completed their detailed design.Secondly, I have gathered some sample data from Ganzhou Teleco mmunication Corporation, confirmed the system by some empirical values, and carried on the comparative analysis to the segmentation result.The experimental result indicated that segmentation results from tt-k-means and tt-k-mediods algorithms were efficient and effective, and the model produced was reasonable and understandable.Finally, I introduced the entire flow of data mining and the framework of data warehouse, and completed the customer segmentation system in the construction. The system took tt(two-times) algorithms as a core. It’s composed of system administration module, pretreatment module, cluster analysis module, knowledge expression module. This system should become a subsystem to the business analysis system.The successful application of this system would facilitate to explore some other similar segmentation system.更多还原