基于数据挖掘的电信领域客户流失研究-软件工程专业论文.docxVIP

摘要 数据挖掘是从大量的、不完全的、模糊的、随机的数据中，提取隐含在其中 的、人们事先不知道的、但又是潜在有用的信息和知识的过程。数据挖掘过程一 般包含业务理解、数据准备、建棋、以及评估等几个步骤。 数据挖掘算法经过多年的明究现在已经基本成熟和稳定，数据挖掘的重心正 在转向各个应用领域，在电仁、 领域内具体实施商业数据挖掘时需要对电信业务、 流程做到全面理解。 本文首先简要介绍数据挖掘概念和基本技术，探讨了几个标准数据挖掘流 程。在总结己经实施过的电伯领域数据挖掘过程的经验基础上提出电信领域内决 策分析流程的商业理解、!扎!，Jlli fiJN-和数据准备阶段流程 2 先以宏观角度理解一般 的电信领域内业务知识和|均业流利，从企业经营基本问题获取利润出发， 采取一个设问的流程将大的 ji 业问题分解为若干子问题，并不断继续分解，获得 基本的商业问题集，然后在汇集回答这些商业问题所需要分析因素;在具体数据 集上得到分析因素到物理表和手段的映射关系。最后以电信领域客户流失为例实 施数据挖掘过程。 关键词:数据挖掘、客户流失、 CRISP-DM 、商业理解 Abstract Data mining is the process 0 f revealing instructive but buried information and knowledge hidden in a large. incol11plete ，fuzzy，stochastic data. It always includes business understanding ，data pr 叩arat lO口， model building and results evaluation After several years research clata mining algorithrns is now mat田e and steady. The focus has turnecl to applicalion clol11ain. In telecommunication industry，when implementing data mining process analyst should fully understand business and process ofthe domain. This paper introclllces basic concepts and technology of data mining，and discusses several standard data l11ining process model. Based on experience from practical mining work in teleco l11l11l1n ?cation clomain，we present the process of business understandi 吨， data lInderstanding. data preparation in data mining in 也e domain. First of all，cOl11prehencl the 吕.:neral bllsiness knowledge and process in the teleco domain. From the essential qucstion in bllsiness ，how to gain profit，we practice a questioning process: divide ;1 gen巳ral qlles!ìon into several sub-questions，and continue divide sllb-qllestions lI ntil acqllire a collection ofbusiness questions; sum up analytical factors; strike lIP the mapping relationship between the factors and physical data such as table and field in database. In the latter p缸t of this thesis，a process of mining churn in teleco indllstry is implemented. Key words: data mining，Ch1ll1l. CRISP-DM ，bllsiness underst