聚类集成算法在客户细分模型中的应用  

Application of Clustering Integration Algorithm in Customer Segmentation Model

作　　者：黄飞杰[1] 郎旭明 尹健康[3] 张卫东[4] 唐靖 HUANG Feijie;LANG Xuming;YIN Jiankang;ZHANG Weidong;TANG Jin(Internet Research Center,Guangxi Zhongyan Industry Co.,Ltd.,Nanning 530000,Guangxi,China;Hebei Provincial tobacco company,Cangzhou company,Cangzhou 061001,Hebei,China;Sichuan Provincial Tobacco Companies Chengdu Company,Chengdu 610072,Sichuan,China;School of Computer Science and Technology,Southwest University of Science and Technology,Mianyang 621010,Sichuan,China;School of Manufacturing Science and Engineering,Southwest University of Science and Technology,Mianyang 621010,Sichuan,China)

机构地区：[1]广西中烟工业有限责任公司互联网研究中心,广西南宁530000 [2]河北省烟草公司沧州市公司,河北沧州061001 [3]四川省烟草公司成都市公司,四川成都610014 [4]西南科技大学计算机科学学院,四川绵阳621010 [5]西南科技大学制造科学与工程学院,四川绵阳621010

出　　处：《西南科技大学学报》2020年第1期75-80,共6页Journal of Southwest University of Science and Technology

基　　金：四川省科技计划资助项目(2017GZ0358,2017GZ0187)。

摘　　要：营销分析、客户分析和内部运营管理是企业大数据应用最广泛的3个领域。基于某市烟草零售数据,运用聚类集成算法构建客户细分模型,实现烟草零售精准营销。本算法集成了Clara算法和K-means算法,首先采用Clara算法随机生成初始聚类中心,K-means算法进行无监督学习,得到最少代价的聚类中心,然后基于零售客户基本信息进行聚类因子建模,将类别与档位进行维度交叉,对零售客户在档位内二次细分,优化迭代分配,解决烟草零售客户档位内部差异化大的问题。经多种聚类算法比较,本文方法的总误差和误差波动率小于其他单一的聚类算法。将该方法用于烟草营销投放策略,经市场验证,能够提升订足率和需求满足率。Marketing analysis,customer analysis and internal operation management are the three most widely used fields of enterprise big data application scenarios.Based on the tobacco retail data of a city,the clusteringintegration algorithm,which integrates Clara algorithm and K-means algorithm,is applied to build a customer segmentation model to achieve tobacco retail precision marketing.Clara algorithm is used to randomly generate initial cluster centers and K-means algorithm is used for unsupervised learning to get the least cost cluster centers.Based on the basic information of tobacco retail customers,a clustering factor model is built.To solve the problem of large internal differentiation of tobacco retail customers,the categories and gears are dimensioned,and the tobacco retail customers are subdivided in the gears to optimize the iterative allocation.After comparing multiple clustering algorithms,the total error and error volatility of the clustering integration algorithm are smaller than those of other single clustering algorithms.Applying this method to the tobacco marketing strategy,the market verification of the products of the two brands can improve the subscription rate and the demand satisfaction rate.

关 键 词：聚类集成算法 客户细分模型 投放策略 需求满足率 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论] F274[自动化与计算机技术—计算机科学与技术]