基于改进K-means算法的电子商务客户细分研究

Research on e-commerce customer segmentation basedon improved K-means clustering

目的:针对传统K-means算法需要人为设定K值的缺陷,提出改进的K-means算法,并将其应用于电子商务客户细分研究。方法:首先,在经典RFM模型的基础上,增加客户消费行为特征;其次,为确定最佳聚类数目,引入CH评价指标,以对K-means算法进行改进;最后,选取了包含37376个样本的电子商务客户数据集进行实证研究。结果:与拐点法相比,通过CH指标确定K值更加直观;与谱聚类相比,加入CH指标的K-means算法具有更优的聚类效果及运行效率。结论:结合CH聚类评价质量指标和K-means算法能有效提高电子商务客户细分的准确性和效率。

Aims:In view of the shortcoming of setting K value artificially in the traditional K-means algorithm,we proposed an improved K-means algorithm.This algorithm was employed to segment e-commerce customers.Methods:Firstly,on the base of the RFM model,the features of customer consumption behavior was introduced.Secondly,in order to determine the optimal number of clustering,the CH index was introduced to improve the K-means algorithm.Finally,a dataset with 37,376 samples of e-commerce customers was selected for empirical research.Results:Compared with the inflection point method,K values selected by the CH index was more intuitive.Compared with the spectral clustering,the clustering effect and operating efficiency of the K-means algorithm with the CH index were better.Conclusions:The accuracy and efficiency of e-commerce customer segmentation could be improved effectively by combining the CH index and the K-means algorithm.