摘要
通过对电信业客户流失预测的国内外研究成果的分析,我们发现造成电信业客户流失原因种类比较多、难以用一种通用的划分标准对流失客户的流失特征进行刻画,因此本文提出了将K-means算法与传统的分类算法相结合的方法进行客户流失分析,并进行了应用实验.该实验以中国联通湖南某地区X分公司的客户数据为基础,利用数据挖掘软件Clementine8.1建立了客户流失分类预测模型,模型的应用结果表明:新方法对客户流失预测的命中率高于传统的分类预测算法.
Through the analysis of inland cause of the customer prediction for churning in and overseas research results, it is discovered that the telecom is various and it＇s difficult to describe the characteristics of churning customers in a general division standard. This article presented a method combining K-means with classification predicting algorithm to analyze the characteristics of churning customers, and the application experiment was carried out based on the customers＇ data from X subsidiary company of China Unicorn in Hunan, using data mining software Clementine 8.1, to establish the prediction model for churning. The result shows that the predicting hit rate of the new method is obviously higher than that of traditional classification predicting algorithm.
出处
《佳木斯大学学报（自然科学版）》
CAS
2010年第2期175-179,共5页
Journal of Jiamusi University：Natural Science Edition
基金
福建工程学院院基金项目(GY-Z08102)
关键词
电信
客户流失
流失预测模型
telecom
customer churning
prediction model for churning
