改进的K-means算法在电信客户细分中的应用  

Application of Improved K-means Algorithm in Subdivision of Telecom Clients

作　　者：耿筱媛[1] 张燕平[1] 闫屹[1] 

机构地区：[1]安徽大学计算智能与信号处理教育部重点实验室,安徽合肥230039

出　　处：《计算机技术与发展》2008年第5期163-167,共5页Computer Technology and Development

基　　金：国家自然科学基金项目(60675031;60475017);973计划(国家重点基础研究)(2004CB318108);安徽省教育厅重点自然科学研究项目(2006kj015A);安徽省教育厅自然科学研究项目(2005kj053);安徽大学211工程学术创新团队

摘　　要：在K-means算法中,选择不同的初始聚类中心会产生不同的聚类结果且有不同的准确率,并且其迭代过程在时间上不是高效的。针对K-means算法的这两点不足做了一定程度上的改进,理论分析表明,改进后的算法具有较高的准确度和较低的时间复杂度。采用改进后K-means聚类算法对电信客户数据进行聚类分析,得到具有不同特征的客户群组,通过与统计分析的对比,聚类结果分析更合理清晰,更便于对不同群组采取不同的经营策略,为管理者提供了合理的决策支持。In the K- means algorithm, selecting different initial centers that the algorithm begins with can produce different results and different accuracy,and the process of iterative is not high - efficiency. Investigates the standard K - means clustering algorithm and gives an improved algorithm by selecting better initial centers that the algorithm begins with, and improved its efficiency. The theoretic analysis shows that the improved K - means algorithm can get higher accuracy and better time - complexity. The application of this improved K- means algorithm in subdivision of telecorn clients can improve the relationship between the organization and the customers and forecast the trend and behaviors to support people＇ s decision. K - means algorithm is used to analyze sample data and discover varied characters of varied groups of customers. It can support enterprise to make an appropriate decision for each kind of customer.

关 键 词：数据挖掘 聚类算法 K-MEANS算法 准则函数 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]