基于数据挖掘的零售客户细分模型的应用研究
|
摘要
随着信息网络的迅猛发展,零售市场以客户为中心的、服务为目的的战略显得至关重要,掌握客户的需求趋势,加强与客户的关系,有效发掘、管理客户资源是企业的致胜关键。因此,客户关系管理成为了零售领域的研究焦点。客户细分是客户关系管理中的首要任务,只有得到好的细分模型才能有效地对客户进行识别,进而客户保持、客户吸引才能针对性地实施。
许多营销者都相信,行为是构建细分市场的最佳起点,本文选择了基于客户行为的细分方法。对于细分变量的选取,能够兼顾客户价值和客户关系的质量最好,然而现实中却很难做到。基于行为的细分中,经典的RFM分析、客户价值矩阵分析虽然都是相对有效的细分方法,但是都忽略了一个关键因素,那就是客户忠诚,忠诚客户群带来的盈利对企业也是很重要的。因此,本文在价值矩阵分析方法的基础上,选取了平均购买额和购买频率两个变量之后,增加了一个用以表征客户存在时间的变量,即客龄,可以在一定程度上体现客户的稳定性。
本文利用数据挖掘技术,以平均购买金额、购买频率和客龄作为细分变量,构建零售业的客户细分模型。客户聚类,K-均值算法是一种常用算法,但是聚类数目要预先指定,初始聚类质心是随机选取的,聚类效果未必令人满意。自组织神经网络SOM算法,能够自适应的将样本数据划分成不同的类,不需要预先设定聚类数目,但是不能提供分类后精确的聚类信息。于是,本文提出一种将两算法结合的方式,把整个聚类分析分为了两阶段进行:第一阶段使用SOM神经网络得到聚类数目与聚类质心;第二阶段用第一阶段的输出作为k-均值算法的输入。
将客户分为不同类别后,对每一类客户的特征进行提取,有助于提高营销活动的针对性和有效性,有助于客户关系管理的良好实施,本文应用数据挖掘中的决策树来提取客户特征。最后,进行购买参照分析,分析每一类客户的特征与购买商品之间的关联。客户特征提取和购买参照即是对所建客户细分模型的应用。
Along with the rapid development of information network, the strategy centering on customer and aiming on service in retailing are becoming important. It is the key that successfully mastering the trend of customers’requirement, strengthening the connection with them, efficiently digging and managing their information. Thus, the customer relation management becomes the focus in the research of retail domain. The customer subdivision is the chief task of the customer relation management. Only the good subdivision model is build, then the customers can be effectively identified, and the customer keeping and the customer attracting can be put into practice.
A lot of sellers believe that behavior is the best jumping-off point in building a customer subdivision model, the paper chooses the method which basing on customer behavior to build a subdivision model. For selecting subdivision variable, the customer’s value and quality of relationship should be considerate simultaneously; however, it is hard to live up to. Among the behavior-based subdivision methods, although both the classical RFM analysis, and customer value matrix analysis are relatively efficient subdivision methods, they neglect a point, that’s the customer’s loyalty, the profit brought by loyal customers is also very important for the corporation. Hence, after selecting the average monetary and the frequency of purchase basing on the value matrix analysis, another variable is selected to embody the stability of customer’s existing time.
In this paper, the data mining technology is used, the average monetary and frequency of purchase and existing time is used as the behavior variables to build the retailing customer subdivision model. The k-means algorithmic is widely used in clustering analysis, but the number of clustering needs to be appointed by beforehand, and the initial clustering centers is selected in random, the last clustering effect is turn up trumps unnecessarily. The SOM (self-organizing feature map) neural networks arithmetic divides the sample data into different clusters by self adaptively, and the clustering number is needn’t to be pre-established, but the accurate clustering information can not be offered. Upon then, a combinative arithmetic is brought forward. The clustering analysis process is composed of two steps: At the first step, using SOM neural networks to get clustering number and the clustering centers; at the second step, the output of the first step is used as input of the second step which is the k-means.
After dividing the customers into different classes, picking up the characteristics of each class of customers, that is helpful for boosting the pertinence and the validity of sell activity, help for bring into effect of customer relation management, the decision tree in data mining is used for picking up customers’characteristics. At last, purchase consult analysis is carried through, that is analyzing the association of customers’characteristics and the commodities purchased. Pick-up customers’characteristics and purchase consult analysis are the application of the customer subdivision model just constructed.
许多营销者都相信,行为是构建细分市场的最佳起点,本文选择了基于客户行为的细分方法。对于细分变量的选取,能够兼顾客户价值和客户关系的质量最好,然而现实中却很难做到。基于行为的细分中,经典的RFM分析、客户价值矩阵分析虽然都是相对有效的细分方法,但是都忽略了一个关键因素,那就是客户忠诚,忠诚客户群带来的盈利对企业也是很重要的。因此,本文在价值矩阵分析方法的基础上,选取了平均购买额和购买频率两个变量之后,增加了一个用以表征客户存在时间的变量,即客龄,可以在一定程度上体现客户的稳定性。
本文利用数据挖掘技术,以平均购买金额、购买频率和客龄作为细分变量,构建零售业的客户细分模型。客户聚类,K-均值算法是一种常用算法,但是聚类数目要预先指定,初始聚类质心是随机选取的,聚类效果未必令人满意。自组织神经网络SOM算法,能够自适应的将样本数据划分成不同的类,不需要预先设定聚类数目,但是不能提供分类后精确的聚类信息。于是,本文提出一种将两算法结合的方式,把整个聚类分析分为了两阶段进行:第一阶段使用SOM神经网络得到聚类数目与聚类质心;第二阶段用第一阶段的输出作为k-均值算法的输入。
将客户分为不同类别后,对每一类客户的特征进行提取,有助于提高营销活动的针对性和有效性,有助于客户关系管理的良好实施,本文应用数据挖掘中的决策树来提取客户特征。最后,进行购买参照分析,分析每一类客户的特征与购买商品之间的关联。客户特征提取和购买参照即是对所建客户细分模型的应用。
Along with the rapid development of information network, the strategy centering on customer and aiming on service in retailing are becoming important. It is the key that successfully mastering the trend of customers’requirement, strengthening the connection with them, efficiently digging and managing their information. Thus, the customer relation management becomes the focus in the research of retail domain. The customer subdivision is the chief task of the customer relation management. Only the good subdivision model is build, then the customers can be effectively identified, and the customer keeping and the customer attracting can be put into practice.
A lot of sellers believe that behavior is the best jumping-off point in building a customer subdivision model, the paper chooses the method which basing on customer behavior to build a subdivision model. For selecting subdivision variable, the customer’s value and quality of relationship should be considerate simultaneously; however, it is hard to live up to. Among the behavior-based subdivision methods, although both the classical RFM analysis, and customer value matrix analysis are relatively efficient subdivision methods, they neglect a point, that’s the customer’s loyalty, the profit brought by loyal customers is also very important for the corporation. Hence, after selecting the average monetary and the frequency of purchase basing on the value matrix analysis, another variable is selected to embody the stability of customer’s existing time.
In this paper, the data mining technology is used, the average monetary and frequency of purchase and existing time is used as the behavior variables to build the retailing customer subdivision model. The k-means algorithmic is widely used in clustering analysis, but the number of clustering needs to be appointed by beforehand, and the initial clustering centers is selected in random, the last clustering effect is turn up trumps unnecessarily. The SOM (self-organizing feature map) neural networks arithmetic divides the sample data into different clusters by self adaptively, and the clustering number is needn’t to be pre-established, but the accurate clustering information can not be offered. Upon then, a combinative arithmetic is brought forward. The clustering analysis process is composed of two steps: At the first step, using SOM neural networks to get clustering number and the clustering centers; at the second step, the output of the first step is used as input of the second step which is the k-means.
After dividing the customers into different classes, picking up the characteristics of each class of customers, that is helpful for boosting the pertinence and the validity of sell activity, help for bring into effect of customer relation management, the decision tree in data mining is used for picking up customers’characteristics. At last, purchase consult analysis is carried through, that is analyzing the association of customers’characteristics and the commodities purchased. Pick-up customers’characteristics and purchase consult analysis are the application of the customer subdivision model just constructed.
引文
[1] 李莉.会员卡营销:我国零售企业建立客户忠诚的有力武器[J].商场现代化,2005,8:32-33
[2] 张红梅,夏南强.数据挖掘技术在零售业中的应用[J].经济师,2006.1:48-49
[3] 汪旭晖.外资企业进入中国市场的现状分析[J].商业经济文荟,2005.6:13-16
[4] 王伟民.入世后过渡期我国零售业的现状及发展对策[J].世界贸易组织动态与研究,2005.6:12-14
[5] 刘凌.运用 CRM 策略加强品牌建设和市场营销——零售业 CRM 策略研究[EB/OL]. http://www.amteam.org/
[6] ZhaoHui Tang,Jamie MacLennan.Data Mining with SQL Server 2005[M].Wiley Publishing,2005:5-8
[7] 林宇.数据仓库原理与实践[M].北京:人民邮电出版社,2003:123-124
[8] Data mining with Clementine for smarter retailing[Z].SPSS White Paper
[9] 迈克尔·J·A·贝里,戈登·S·利诺夫著.袁卫等译.数据挖掘:客户关系管理的科学与艺术[M].北京:中国财政经济出版社,2003.8:285,303-308
[10] Mark F. Hornick, Erik Marcadé, Sunil Venkayala.Java Data Mining: Strategy, Standard, and Practice: A Practical Guide for architecture, Design, and Implementation[M].2006
[11] CRISP-DM 1.0 数据挖掘指导手册.CRISP-DM 委员会,2000:7-8
[12] 邵峰晶,于忠清.数据挖掘原理与算法[M].北京:中国水利水电出版社,2003:130-131
[13] 齐佳音,万映红.客户关系管理理论与方法[M].北京:中国水利水电出版社;知识产权出版社,2006:2-3
[14] 胡运发.数据与知识工程导论[M].北京:清华大学出版社,2003:135-136,144-145
[15] (美)贝尔森,(美)史密斯,(美)西瑞林.构建面向 CRM 的数据挖掘应用[M].人民邮电出版社,2001.8:167-168
[16] (加)韩家炜,(加)坎伯著;范明等译.数据挖掘概念与技术[M].机械工业出版社,2001.8:303-304
[17] 叶孝明.数据挖掘在客户关系管理中的应用研究[J].东南大学,2006.3:28
[18] 刘英姿,何伟.基于不同视角的客户细分方法研究综述[J].商场现代化,2007,1:271-271
[19] Olivia Parr Rud.DataMining Cookbook Modeling Data for Marketing, Risk, and Customer Relationship Management[M].Wiley Computer Publishing, 2001:191-193
[20] 张云涛,龚玲.数据挖掘原理与技术[M].北京:电子工业出版社,2004.4:10-16
[21] 陈伟,陈耿.基于业务规则的错误数据清理方法[J].计算机工程与应用,2005:172-714
[22] 陈伟,丁秋林.数据清理中不完整数据的清理方法[J].微型机与应用,2005.2:44-45
[23] 赵群.消除重复数据清理方法的研究[J].福建电脑,2006.4:51-52
[24] 飞思科技产品研发中心编著.SQL Server 2000 数据库和数据仓库[M].北京:电子工业出版社,2001.6:333-338
[25] 康塔尼克著,闪四清等译.数据挖掘:概念、模型、方法和算法[M].北京:清华大学出版社,2003 :30,146-149
[26] Pang-Ning Tan,Micheal Steinbach,Vipin Kumar.《数据挖掘导论》Introduction to Data Mining[M].人民邮电出版社,2006,5:101-102,307
[27] Michael J. A. Berry, Gordon S. Linoff.Data Mining Techniques: For Marketing, Sales, and Customer Relationship Management[M].Second Edition. Wiley Computer Publishing, 2004:111-113
[28] 刘英姿,吴昊.客户细分方法研究综述[J].管理工程学报,2006,20(1):271
[29] 叶孝明,黄祖庆.基于数据挖掘的零售业客户细分研究[J].现代管理科学,2006.6:63-64
[30] 贺蓓.基于行为表现的客户忠诚评价指标分析[J].商场现代化,2006.5:179
[31] 李卫东,张桂芝,李欣.利用数据挖掘方法分析客户忠诚度[J].计算机与网络,2005.2:59-61
[32] 吴靖,林政.基于持卡消费行为的客户细分模型研究[J].中南财经大学学报,2006.6:67-68
[33] (美)米哈尔斯基著,朱明等译.机器学习与数据挖掘:方法和应用[M].北京:电子工业出版社,2004.1:40-42
[34] 谭元戎,孙剑平.聚类模型在客户关系管理中的应用[J].技术经济,2007,26(5):51-54
[35] 陈莉,焦李成.基于自适应聚类的数据预处理算法[J].计算机应用与软件,2005.3,22(3):28-29
[36] 阎平凡,张长水.人工神经网络与模拟进化计算[M].北京:清华大学出版社,2005:215-219
[37] 谷萩隆嗣,萩原将文等著.马炫译.人工神经网络与模糊信号处理[M].北京:科学出版社,2003:41-43
[38] 陈泯融,邓飞其.基于 SOM 网络的数据聚类[J].电脑知识与技术,2004.11:59-60
[39] 杨黎刚,苏宏业,张英.基于 SOM 聚类的数据挖掘方法及其应用研究[J].计算机工程与科学,2007.11,29(8):133-136
[40] 陈伯成,梁冰,周越博.自组织映射神经网络(SOM)在客户分类中的一种应用[J].系统工程理论与实践,2004.3:8-14
[41] 刘林,喻国平.基于自组织特征映射(SOM)网络对潜在客户的挖掘[J].南昌大学学 报,2006.6,30(5):507-510
[42] 杨黎刚.基于 SOM 聚类的数据挖掘方法及其应用研究[D].浙江大学硕士学位论文, 2006.8:35-36
[43] Daniel T. Larose.Discovering Knowledge in Data: An Introduction to DataMining[M]. Wiley Computer Publishing, 2005:153, 166-170
[44] 刘同明.数据挖掘技术及其应用[M].北京:国防工业出版社,2001.9:124-126
[45] Xia Shixiong,Li Wenchao.Improved k-means clustering algorithm[J].Journal of Southeast University (English Edition),2007,23(9):435-438
[46] 汉德著.张银奎等译.数据挖掘原理[M].北京:机械工业出版社,2003.4:191-192
[47] 邢留伟.k-means 算法在客户细分中的应用研究[D].西南财经大学硕士学位论文,2004.4:17
[48] 周欢,黄立平.基于 SOM 神经网络的 C-均值聚类算法[J].计算机应用,2007,27(6):51-52
[49] 杨占华,杨燕.一种基于 SOM 和 K-means 的文档聚类算法[J].计算机应用研究,2006.5:73-74
[50] 苏克,戴维森著.朱建秋,蔡伟杰译.可视化数据挖掘:数据挖掘可视化和数据挖掘的技术与工具[M].北京:电子工业出版社,2004.1:71-76
[51] 鲁为,王枞.决策树算法的优化与比较[J].计算机工程,2007.8,33(16):189-190
[52] 沈晨鸣.决策树分类算法研究[J].盐城工学院学报,2005.12,18(4):22-24
[53] 杨善林,倪志伟.机器学习与智能决策支持系统[M].北京:科学出版社,2004:71-72
[54] 唐敏.关联规则挖掘算法在超市销售分析中的应用[J].计算机科学,2006,33(2):149-150
[55] 左映华,高居泰,李晋宏.Carma 关联规则算法的一种改进[J].韶关学院学报(自然科学版),2004.3,25(3):51
[56] 李大勇,徐亚娜.CRM 中 Apriori 算法的应用研究[J].辽宁交通高等专科学校学报,2007.6,9(2):87-89
[57] (美)林那夫,(美)密歇尔著;沈钧毅等译.Web 数据挖掘:将客户数据转化为客户价值[M].电子工业出版社,2004.3:51
[58] David Taniar.Research and Trends in Data Mining Technologies and Applications[M]. Published in the United States of America by Idea Group Publishing, 2007:121
[2] 张红梅,夏南强.数据挖掘技术在零售业中的应用[J].经济师,2006.1:48-49
[3] 汪旭晖.外资企业进入中国市场的现状分析[J].商业经济文荟,2005.6:13-16
[4] 王伟民.入世后过渡期我国零售业的现状及发展对策[J].世界贸易组织动态与研究,2005.6:12-14
[5] 刘凌.运用 CRM 策略加强品牌建设和市场营销——零售业 CRM 策略研究[EB/OL]. http://www.amteam.org/
[6] ZhaoHui Tang,Jamie MacLennan.Data Mining with SQL Server 2005[M].Wiley Publishing,2005:5-8
[7] 林宇.数据仓库原理与实践[M].北京:人民邮电出版社,2003:123-124
[8] Data mining with Clementine for smarter retailing[Z].SPSS White Paper
[9] 迈克尔·J·A·贝里,戈登·S·利诺夫著.袁卫等译.数据挖掘:客户关系管理的科学与艺术[M].北京:中国财政经济出版社,2003.8:285,303-308
[10] Mark F. Hornick, Erik Marcadé, Sunil Venkayala.Java Data Mining: Strategy, Standard, and Practice: A Practical Guide for architecture, Design, and Implementation[M].2006
[11] CRISP-DM 1.0 数据挖掘指导手册.CRISP-DM 委员会,2000:7-8
[12] 邵峰晶,于忠清.数据挖掘原理与算法[M].北京:中国水利水电出版社,2003:130-131
[13] 齐佳音,万映红.客户关系管理理论与方法[M].北京:中国水利水电出版社;知识产权出版社,2006:2-3
[14] 胡运发.数据与知识工程导论[M].北京:清华大学出版社,2003:135-136,144-145
[15] (美)贝尔森,(美)史密斯,(美)西瑞林.构建面向 CRM 的数据挖掘应用[M].人民邮电出版社,2001.8:167-168
[16] (加)韩家炜,(加)坎伯著;范明等译.数据挖掘概念与技术[M].机械工业出版社,2001.8:303-304
[17] 叶孝明.数据挖掘在客户关系管理中的应用研究[J].东南大学,2006.3:28
[18] 刘英姿,何伟.基于不同视角的客户细分方法研究综述[J].商场现代化,2007,1:271-271
[19] Olivia Parr Rud.DataMining Cookbook Modeling Data for Marketing, Risk, and Customer Relationship Management[M].Wiley Computer Publishing, 2001:191-193
[20] 张云涛,龚玲.数据挖掘原理与技术[M].北京:电子工业出版社,2004.4:10-16
[21] 陈伟,陈耿.基于业务规则的错误数据清理方法[J].计算机工程与应用,2005:172-714
[22] 陈伟,丁秋林.数据清理中不完整数据的清理方法[J].微型机与应用,2005.2:44-45
[23] 赵群.消除重复数据清理方法的研究[J].福建电脑,2006.4:51-52
[24] 飞思科技产品研发中心编著.SQL Server 2000 数据库和数据仓库[M].北京:电子工业出版社,2001.6:333-338
[25] 康塔尼克著,闪四清等译.数据挖掘:概念、模型、方法和算法[M].北京:清华大学出版社,2003 :30,146-149
[26] Pang-Ning Tan,Micheal Steinbach,Vipin Kumar.《数据挖掘导论》Introduction to Data Mining[M].人民邮电出版社,2006,5:101-102,307
[27] Michael J. A. Berry, Gordon S. Linoff.Data Mining Techniques: For Marketing, Sales, and Customer Relationship Management[M].Second Edition. Wiley Computer Publishing, 2004:111-113
[28] 刘英姿,吴昊.客户细分方法研究综述[J].管理工程学报,2006,20(1):271
[29] 叶孝明,黄祖庆.基于数据挖掘的零售业客户细分研究[J].现代管理科学,2006.6:63-64
[30] 贺蓓.基于行为表现的客户忠诚评价指标分析[J].商场现代化,2006.5:179
[31] 李卫东,张桂芝,李欣.利用数据挖掘方法分析客户忠诚度[J].计算机与网络,2005.2:59-61
[32] 吴靖,林政.基于持卡消费行为的客户细分模型研究[J].中南财经大学学报,2006.6:67-68
[33] (美)米哈尔斯基著,朱明等译.机器学习与数据挖掘:方法和应用[M].北京:电子工业出版社,2004.1:40-42
[34] 谭元戎,孙剑平.聚类模型在客户关系管理中的应用[J].技术经济,2007,26(5):51-54
[35] 陈莉,焦李成.基于自适应聚类的数据预处理算法[J].计算机应用与软件,2005.3,22(3):28-29
[36] 阎平凡,张长水.人工神经网络与模拟进化计算[M].北京:清华大学出版社,2005:215-219
[37] 谷萩隆嗣,萩原将文等著.马炫译.人工神经网络与模糊信号处理[M].北京:科学出版社,2003:41-43
[38] 陈泯融,邓飞其.基于 SOM 网络的数据聚类[J].电脑知识与技术,2004.11:59-60
[39] 杨黎刚,苏宏业,张英.基于 SOM 聚类的数据挖掘方法及其应用研究[J].计算机工程与科学,2007.11,29(8):133-136
[40] 陈伯成,梁冰,周越博.自组织映射神经网络(SOM)在客户分类中的一种应用[J].系统工程理论与实践,2004.3:8-14
[41] 刘林,喻国平.基于自组织特征映射(SOM)网络对潜在客户的挖掘[J].南昌大学学 报,2006.6,30(5):507-510
[42] 杨黎刚.基于 SOM 聚类的数据挖掘方法及其应用研究[D].浙江大学硕士学位论文, 2006.8:35-36
[43] Daniel T. Larose.Discovering Knowledge in Data: An Introduction to DataMining[M]. Wiley Computer Publishing, 2005:153, 166-170
[44] 刘同明.数据挖掘技术及其应用[M].北京:国防工业出版社,2001.9:124-126
[45] Xia Shixiong,Li Wenchao.Improved k-means clustering algorithm[J].Journal of Southeast University (English Edition),2007,23(9):435-438
[46] 汉德著.张银奎等译.数据挖掘原理[M].北京:机械工业出版社,2003.4:191-192
[47] 邢留伟.k-means 算法在客户细分中的应用研究[D].西南财经大学硕士学位论文,2004.4:17
[48] 周欢,黄立平.基于 SOM 神经网络的 C-均值聚类算法[J].计算机应用,2007,27(6):51-52
[49] 杨占华,杨燕.一种基于 SOM 和 K-means 的文档聚类算法[J].计算机应用研究,2006.5:73-74
[50] 苏克,戴维森著.朱建秋,蔡伟杰译.可视化数据挖掘:数据挖掘可视化和数据挖掘的技术与工具[M].北京:电子工业出版社,2004.1:71-76
[51] 鲁为,王枞.决策树算法的优化与比较[J].计算机工程,2007.8,33(16):189-190
[52] 沈晨鸣.决策树分类算法研究[J].盐城工学院学报,2005.12,18(4):22-24
[53] 杨善林,倪志伟.机器学习与智能决策支持系统[M].北京:科学出版社,2004:71-72
[54] 唐敏.关联规则挖掘算法在超市销售分析中的应用[J].计算机科学,2006,33(2):149-150
[55] 左映华,高居泰,李晋宏.Carma 关联规则算法的一种改进[J].韶关学院学报(自然科学版),2004.3,25(3):51
[56] 李大勇,徐亚娜.CRM 中 Apriori 算法的应用研究[J].辽宁交通高等专科学校学报,2007.6,9(2):87-89
[57] (美)林那夫,(美)密歇尔著;沈钧毅等译.Web 数据挖掘:将客户数据转化为客户价值[M].电子工业出版社,2004.3:51
[58] David Taniar.Research and Trends in Data Mining Technologies and Applications[M]. Published in the United States of America by Idea Group Publishing, 2007:121
© 2004-2018 中国地质图书馆版权所有 京ICP备05064691号 京公网安备11010802017129号 地址：北京市海淀区学院路29号 邮编：100083 电话：办公室：(+86 10)66554848；文献借阅、咨询服务、科技查新：66554700 |