【作者】 马晓峰；

【导师】 丁照宇；

【作者基本信息】 成都理工大学 ， 计算机软件与理论， 2011， 硕士

【摘要】 随着电信企业的重组和3G牌照的发放,中国电信市场进入更加激烈的全业务竞争时代,如何适应日趋激烈的市场竞争环境,成为电信企业的重要问题。电信行业重组,中国移动面临着诸多的压力:如何应对由技术发展日新月异而引起的错综复杂的产业格局?如何经营自主研发的、不太成熟、在国际舞台上支撑力度弱的TD标准?如何应对“一家独大”局面受到的不对称监管的困境?如何经营已经开始的全业务模式?如果还是通过一些传统的、简单的数据统计,对于数据的利用仅限于数据的表层信息,而没有去挖掘数据之间更加深层次的信息,是不可能从如此海量的数据和信息中找到解决复杂问题的规律的。数据挖掘技术是一种功能强大的新技术,它能帮助企业在构建数据仓库中找到最重要的信息。本文利用数据挖掘技术找到个人客户的流动特征和规律,并应用数据挖掘技术来构建个人客户识别模型,主要研究内容有:分析并研究数据挖掘技术在移动个人客户识别模型中的应用;研究并初步实现了个人客户识别模型在移动业务相关领域的应用;针对C4.5可以通过改变样本的权重来处理属性值的缺失,利用C4.5的这种特性,对C4.5算法稍做改进可以得到一个基于代价敏感的变种算法C4.5_cs,并将这种算法应用到个人客户识别模型中;在建立个人客户认别模型时,提出了交往指数和符合率的指标,将这两个指标应用到模型构建中,对应用前后的C4.5_cs算法做了对比分析。本文主要基于中国移动现有的经营分析系统,针对中国移动的现实需求,重点研究讨论了决策树算法,从海量的业务系统数据中,分析挖掘个人客户的流动特征,利用决策树算法C4.5_cs建立个人客户识别模型,并利用模型生成的规则实现了模型的应用;模型通过对用户通话特征及个人信息特征等多种信息的分析挖掘,以个人客户交往圈匹配算法为核心,在用户全生命周期(获取期、成长期、成熟期、衰退期、流失期)之外,找到用户在获取期之前(游离期)的来源以及在流失期之后(离网期)的不同去向,使我们对用户的了解更加深入,从而为业务人员进一步了解客户提供帮助,为进行挽留用户和精准营销提供支持。更多还原

【Abstract】 With the telecom restructuring and the issuance of 3G licenses, China’s telecommunications market into the more intense competition in all business era. how to adapt to the increasingly fierce market competition has become the important issue in terms of the telecom business.Telecom industry restructuring, China Mobile is facing many pressures: how to respond to rapid technological development caused by the intricacies of the industry structure? How to run a self-developed, less mature, in the international arena,weak support for the TD standard? How to deal with "a dominant" situation by the plight of the asymmetric regulation? How to run the forthcoming full-service model?Traditional, simple statistics, use of data only for the surface data information between the data and not to dig more in-depth information, is not possible from such a mass of data and information to find solutions to complex problems in the law .By using data mining techniques to find the flow characteristics of individual clients and the law, and apply data mining techniques to build individual customer identification model, the main contents are: analysis and research of data mining technology in the mobile personal customer identification model in the application; study and preliminary recognition model to achieve the individual customers in the mobile business-related field of application; for C4.5 by changing the weight of the sample to deal with missing attribute values, using C4.5 of this characteristic, slight improvement on the C4.5 algorithm can be A variant algorithm based on cost-sensitive C4.5_cs, and this algorithm is applied to individual clients to identify the model; in the establishment of individual customers do not recognize model proposed index and found the rate of exchange index is applied to these two indicators Model, the before and after its application to do a comparative analysis C4.5_cs algorithm.This article is mainly based on China Mobile’s existing management analysis system for the practical needs of China Mobile, the focus of the discussion of the decision tree algorithm, the business system from the mass of data, analysis of the flow characteristics of individual customers mining using decision tree algorithm C4.5_cscreate a personal customer identification model, and the rules generated using the model to achieve application of the model; model calls on the user characteristics and features of personal information mining and other analytical information to individual customers as the core communication loop matching algorithm, the user’s full of lifecycle (access, growth, maturity, decline, loss of period), the period before to find the user access (free) the source and the drain after a period of (off-grid period) of the different fate, so that our users more in-depth understanding, so as to further understanding of the business staff provide help to retain customers and to provide support for precision marketing.更多还原