基于模糊聚类集成算法的客户细分研究.PDFVIP

30 4 Vo.l 30 No. 4 2011 4 JOURNAL OF INTELLIGENCE Apr. 2011 * 1 1 2 高 伟 贺昌政 蒋晓毅 ( 1. 610064; 2. 48149) 针对客户行为的不确定性和模糊性, 将模糊聚类集成技术应用于 CRM 中的客户细分研究, 以提 客户聚 类的精度以模糊 C均值 ( FCM )算法作为基本的聚类器, 应用模糊 t- 范式对生成的多 个聚类器进行集成, 从而获 得最终的客户聚类结果最后, 在 10 个UCI数据集上进行聚类测试, 结果表明, 基于模糊 t- 范式的模糊聚类集成 方法的聚类精度要 于常用的客户聚类 FCM 和 K- m eans方法在客户信用卡数据集 A stralian上的学习曲线还 表明, 聚类集成方法具有更稳定的聚类性能 客户细分 模糊聚类 非监督学习 聚类集成 F272. 1 F713. 50 A 1002- 1965( 2011) 04- 0125- 04 Custom er Segm en tation Study B a sed on Fuzzy Clustering En sem b le 1 1 2 GAO W ei HE Changzheng JIANG X iaoyi ( 1. B siness Schoo,l S ich an Un iversity, Chengd 610064; 2. D epartm ent of M athem atics and Com p ter Science, U niversity ofM nster, M nster 48149, G erm any) A bs tract Apply ing f zzy ensem ble cl ster to c stom er segm entation inCRM in order to im prove the cl ster perfo rm ance ow ing to the n certainty and f zzinesso f c stom er behav io r. First, regarding f zzy C - m eansa lgorithm ( FCM ) asbasic cl sterer, and then sing a f zzy t- no rm to com bine the m ltiple cl sterings to obtain the final consens s cl stering. F inally, experim ents on 10 UC I data sets show that the f zzy cl stering ensem ble algorithm based on t- No rm is better than FCM and K - m eans. F rtherm ore, expermi ents on A stralian data set also indicate that the cl stering ensem ble algorithm w e propo sed has better stability. K ey word s c stom er segm entation f zzy cl ster ns perv ised learning ensem ble cl stering