主管：中国科学院
主办：中国优选法统筹法与经济数学研究会
　　　中国科学院科技战略咨询研究院
中国管理科学

中国管理科学

• 论文 • 上一篇    下一篇

基于OSA算法和GMDH网络集成的电子商务客户流失预测

朱帮助1,2, 张秋菊1,2, 邹昊飞3, 魏一鸣2   

  1. 1. 五邑大学经济管理学院, 广东 江门 529020;
    2. 北京理工大学管理与经济学院, 北京 100081;
    3. 中国国际工程咨询公司, 北京 100048
  • 收稿日期:2009-12-04 修回日期:2011-07-12 出版日期:2011-10-30 发布日期:2011-10-30
  • 作者简介:朱帮助(1979- ),男(汉族),江苏宿迁人,五邑大学经济管理学院副教授,北京理工大学管理与经济学院博士后,研究方向:复杂系统分析与建模、CRM建模、智能信息处理理论与应用研究
  • 基金资助:
    国家自然科学基金资助项目(70471074);国家博士后科学基金资助项目(20100470008);广东省自然科学基金资助项目(9452902001004060)

E-Business Customer Churn Prediction Based on Integration of Objective System Analysis and Group Method of Data Handling Network

ZHU Bang-zhu1,2, ZHANG Qiu-ju1,2, ZOU Hao-fei3, WEI Yi-ming2   

  1. 1. School of Economics and Management, Wuyi University, Jiangmen 529020, China;
    2. School of Management and Economics, Beijing Institute of Technology, Beijing 100081, China;
    3. China International Engineer Consulting Corporation, Beijing 100048, China
  • Received:2009-12-04 Revised:2011-07-12 Online:2011-10-30 Published:2011-10-30

PDF (PC)

2671

被引次数

11

摘要： 电子商务客户流失预测是一种典型的高维、非线性、数据不平衡问题,传统的方法已很难提高其预测精度。本文将自组织数据挖掘方法(SODM)引入电子商务客户流失预测,提出一种基于客观系统分析(OSA)和数据分组处理(GMDH)网络集成的电子商务客户流失预测模型。首先利用OSA算法自动选择出重要的电子商务客户流失关键属性,然后将训练样本送入GMDH网络进行学习与训练,进而对测试样本客户流失状态进行预测。为了提高预测精度,本文还利用向上采样法进行数据平衡化,使得流失类和非流失类客户数量大致相等。应用该模型对某网上商场客户流失状态进行预测,并将预测结果与神经网络、SVM等方法得到的结果进行了比较,验证了该模型的有效性及实用性。

关键词: 自组织数据挖掘, 客观系统分析, 数据分组处理, 客户流失预测, 电子商务

Abstract: Facing with the high dimensional,nonlinear and unbalanced data problems of churn prediction of E-business customers,it is difficult to improve the accuracy of churn prediction of E-business customers by applying traditional methods.Hence an integration model for churn prediction of E-business customers based on objective system analysis (OSA) and group method of data handling (GMDH),two important selforganized data mining (SODM) algorithms,is presented in this paper.Firstly,the key attributes are automatically selected using OSA algorithm.Then GMDH network is trained with training samples,which is used to identify customer churn status of testing samples.Up-sampling metod is also used in this paper to balance the churn-customer data and unchurn-customer data to improve the forecasting accuracy.This proposed approach is applied for chum prediction of an online shop,which proves that compared with some common approaches such as artificial neural networks and support vector machines,more accuracy forecasted results can be obtained.

Key words: self-organized data mining, objectire system analysis, group method of data mining, churn pre-diction, E-business

中图分类号: