联系我们

论文查询结果	返回搜索

论文编号：	1639
作者编号：	2120082046
上传时间：	2010/6/13 1:29:15
中文题目：	基于粗糙集的直销行业客户分类研究
英文题目：	Direct-Selling Customer Classification Study Based on Rough Set
指导老师：	安利平
中文关键字：	粗糙集、客户分类、客户关系管理、数据挖掘
英文关键字：	Rough Set,Customer Classification,Customer Relationships Management,Data Mining
中文摘要：	在激烈的市场竞争中，直销企业发现，只有争取到客户，并且成功做好客户保留，才能在日趋激烈的买方市场竞争环境中使自己生存继而求发展。为了能够了解客户、识别客户，企业开始收集客户信息、分析销售数据，客户分类近年来成为研究的热点，但是随着计算机和网络技术的飞速发展，数据的存储、传输越来越便捷，使得企业囤积了海量的信息，想把这些信息变成有价值的知识并非易事，为此，研究人员做了多方面的尝试，将客户分类与数据挖掘技术相结合来解决这一问题有着现实意义。 现有的基于数据挖掘技术的客户分类研究，多是使用决策树、神经网络、贝叶斯分类等经典的数据挖掘分类方法。但是直销行业以及其他涉及到客户信息的行业，其所拥有的数据中含有大量不确定信息，如空缺值等，由于很多经典的数据挖掘分类算法抗干扰能力有限，使得分类质量大打折扣。客户分类问题的研究中也有少数结合粗糙集理论的，这些也只是利用粗糙集的属性约简、属性权重计算等算法做数据处理，再套用早先的客户分类常用方法如RFM等，但是这些传统的客户分类方法由于属性过少或模型过于简单，也很难建立有效的分类模型解决现实问题。 本文针对直销行业客户数据不完整等特点，遵循CRISP-DM数据挖掘步骤，采用粗糙集理论，经过属性约简、规则导出等算法建立客户分类模型。利用采自我国某直销企业销售部门的客户数据和销售数据完成实证研究。最后将得到的模型与CHAID和QUEST决策树分类模型的分类结果进行比对分析，比较模型分类能力的优劣。
英文摘要：	For the fierce competition,the retailers find that they must persuade customers to buy and make sure to maitain them.Only in this way can they survive.In order to understand and recognize the customers,the retailers began to collect information about customer and sales.But as the computer and network technologies improves,it is so convenient to store or transmit data.As a result the retailers now own a huge amount of information and it is difficult to find valuable knowledge.To solve this problem,researchers tried hard and nowadays customer classification problem become a hot spots.It is valuable to use data mining to sole this problem. Most of the customer classification research uesed some classical data mining technologies such as Decision Tree,Neural Network and Bayes Classification.But in the area of retailing,the data contains many uncertainties such as empty attributes. Because many classical data mining classification algorithms are not good at dealing with information with uncertainty,the model they build is not perfect.Some researchers use Rough Set Theory.But they just use parts of it, like attribute reduction and rule exportation,and then make through it in the old ways like RFM.But these kind of methods take such few attributs into account and make the classification medel not complicated enough to fit the real case. Aim at dealing with the uncertain information,this paper use Rough Set as its theory and build the model with in the steps discussed in CRISP-DM data mining method include the processes of attribute reduction and rule exportation.In order to carry on the empirical study.I use the customer information and sales data from a retailer in our country.And I use three criterions about CRM theory,classfication quality and rule quality to evaluate the model.After that I make a contrast between the Rough Set model andCHAID,QUEST decision tree model,and analyse on which one is more effective in classification.
查看全文：	预览  下载(下载需要进行登录)