数据挖掘在CRM中的应用研究

【作者】 杜松；

【导师】 李小兵；

【作者基本信息】 电子科技大学 ， 机械电子工程， 2004， 硕士

【摘要】 客户关系管理(CRM)是一种旨在改善企业与客户之间关系的新型管理机制,它通过提供优质和个性化的服务吸引和保持更多的客户。经过多年信息管理系统的使用,企业的数据库中拥有关于客户的大量数据,这些数据背后隐藏着许多重要的信息,数据挖掘能够创建对客户进行分类和预测的模型,帮助用户从大量的数据中抽取有用的信息,从而很好地支持企业经营管理者的决策。本文从客户关系管理的需求入手,在分析了数据挖掘的基本原理和技术后,重点讨论了在CRM中应用决策树挖掘技术进行客户分类的问题。为了实现CRM的客户分类,本文作了如下工作:首先,针对SLIQ决策树分类挖掘算法,分析了原算法构造的决策树可能过大,导致用户难以理解和解释,为了限制决策树的大小,对原SLIQ算法提出了一种改进思路。通过输入决策树的最大节点数参数,构造适当大小的树。为了对提出的改进算法进行验证,设计了一个决策树建模软件,验证改进算法的效果,并用该算法对一个客户数据集按客户的个人信息进行客户细分。软件可以实现数据集的可视化表示、挖掘结果的可视化、决策树叶节点规则的提取,能够动态调整决策树建模的输入参数,以便于通过得到的多个不同决策树模型,查看模式的变化情况。最后用PMML规范表示挖掘得到的决策树模型,以方便在不同数据挖掘工具之间共享模型。更多还原

【Abstract】 Customer Relationship Management(CRM) is a kind of new management system aiming at improving the relationship between corporations and their customers. It enables corporation to attract more and more customers through offerring kind and characteristic services. Having use traditional MIS for many years, coporations gathered and stored mass of data about customers. A great deal of important information is concealed under those datasets. Datamining define models which describes customer classification or prediction of customer’s behaviors. Those models help corporations gain useful business information from those data, such kind of information supports corporations to make decisions. Begin with understanding the needs from CRM, this dissertation introduces datamining theroies and techniques at first, then emphasis is placed on the theme how to perform customer segmentation. To abtain the target, works as follows were finished:Firstly, by evaluating problems that too many nodes may be induced by classical SLIQ algorithm even after decision tree pruning, an improved algorithm is proposed. By specifying a parameter of nodes number decision tree at most included, users can get a rough view of the patterns in their data. In order to validate the improved algorithm, a software demo of modeling decision tree is designed and programmed, customer segmentation is executed by this demo according to customers’ information. The demo offer functions as: visualization of datasets and decision tree model, extracting rules from model, analysizing differernt models by adjusting input parameters for building decision tree .Finally, a decision tree model is described by using Predictive Model Markup Language(PMML) document, those kind of documents can be shared amoung differernt dataming tools.更多还原