基于双向聚类的客户细分方法研究

Research on Customer Segmentation Method Based on Biclustering

提出了一种新的客户细分方法。首先,使用卡方统计分析选择属性集并用K-means算法量化每个属性的值。然后,采用基于密度的DBSCAN算法将客户分成三类。最后,使用基于FP-tree的双向聚类算法对三类客户进行聚类获得更详细的信息。以某公司的销售数据集作为实验对象,实验结果表明,双向聚类算法可以更准确、更细致地细分客户。与基于Apriori的双向聚类算法相比,基于FP-tree的双向聚类算法的在大数据集上的效率更高。

This paper proposes a new customer segmentation method.First,selecting a property set using the chi-square statistical analysis and quantify the value of each property using the K-means algorithm.Then,the customer is divided into three classes using the DBSCAN algorithm.Finally,the biclustering algorithm based on FP-tree is used to get more detailed information on three kinds of customers.In this paper,a company＇s sales dataset as the experimental object,experimental results show that the biclustering algorithm can be more accurate and more detailed segmentation of customers.