C均值算法的电信客户细分研究  

Application of Telecom Customer Segmentation Base on Improved C-means Algorithm

作　　者：张焕国[1] 吕莎[1] 李玮[1] 

机构地区：[1]宜宾职业技术学院,四川宜宾644003

出　　处：《计算机仿真》2011年第6期185-188,共4页Computer Simulation

摘　　要：研究准确细分电信客户,提高竞争力,采用随机选择初始值聚类中心和梯度下降寻优方式,易陷入局部最优,导致电信客户细分的准确率低。为了解决C值的不足来提高电信客户细分准确率,提了一种粒子群优化C均值的电信客户细分方法(PSO-FCM)。PSO-FCM通过PSO来选择电信客户细分的初始聚类中心,减小客户细分结果对聚类中心依赖,然后采用C均值算法对电信客户进行细分。在VC++语言环境下,PSO-FCM算法对电信客户消费数据进行仿真,实验结果表明,提高了电信客户细分准确率,更助于电信企业对不同客户群制定相应营销套餐,为电信企业带来更多的收益。C-means algorithm has the initial value sensitive and is a gradient descent algorithm and easily trapped into local optimamum,and customer segmentation accuracy of telecommunications is low.Aiming at the shortcomings of c-means algorithm,telecom customer segmentation method（PSO-FCM） is carried out based on particle swarm optimization and c-means algorithm.PSO-FCM selects the optimal initial c-means clustering center by PSO,and then uses the optimal c-means to solve the telecom customer segmentation problem.In vc＋＋language environment,PSO-FCM is test with a city telecom customer consumption data.The results show that the PSO-FCM has improved the telecom customer segmentation accuracy,and the segmentation results is more scientific,reasonable and more conducive to telecom enterprise for different customer formulate marketing strategy.Tt can improve the marketing effect and quality.

关 键 词：电信客户细分 粒子群算法 聚类分析 

分 类 号：TP311.52[自动化与计算机技术—计算机软件与理论]