SOM聚类方法在卷烟零售户分类中的应用  

Application of SOM Clustering Method to Classification of Cigarette Retailers

作　　者：朱卫东[1] 汪国珍[1] 李矿[2] 

机构地区：[1]合肥工业大学管理学院,合肥市屯溪路193号230009 [2]安徽省淮北市烟草专卖局,安徽省淮北市人民路235000

出　　处：《烟草科技》2014年第6期29-32,共4页Tobacco Science & Technology

摘　　要：为了解卷烟零售户的基本情况、分析其基本特征,根据烟草行业卷烟销售的特点,利用指标组合方法选取零售户经营业态、市场类型等零售户基本属性指标,以及销售行为指标和销售卷烟品种数构建了数据模型,并采用自组织特征映射神经网络(SOM,Self-organizing Feature Maps)方法对零售户进行分类,从最小类间距离和组内离差平方和两方面验证聚类结果,确定最终的零售户分类数。结果表明,将采集的8713个零售户样本划分成7类,分类结果的吻合度最好。根据各聚类簇的特征,从营销服务、市场管理和客户关系管理等方面对各类零售户提出了相应的服务与管理对策。In order to know the basic information of cigarette retailers and analyze their basic characters, on the basis of the nature of tobacco industry, a model was developed with an index combination method by integrating the business status, market type, sale behavior and number of cigarette brand sold. Retailers were clustered with SOM （Self-organizing Feature Maps） method, the results were further verified from the sides of minimum inter-cluster distance and the sum of intra-cluster square dispersion, then the number of clusters for retailers were finally determined. The results showed that the best results were obtained when 8713 retailer samples were grouped into 7 clusters. The corresponding service and management suggestions were provided for the retailers in each group in terms of marketing service, market management, customer relationship management, etc.

关 键 词：客户分类 卷烟零售户 SOM聚类 最小类间距离 数据模型 

分 类 号：F204[经济管理—国民经济]