基于RFM的聚类算法在零售市场客户细分研究

Research on customer segmentation in retail market based on RFM clustering algorithm

客户关系管理作为企业管理的重要组成部分,其客户细分功能直接影响着企业营销战略。为了更好地对零售市场进行客户细分,通过应用某英国零售商数据集,基于RFM模型和4种聚类算法,验证基于RFM模型的K-means、DBSCAN、AGNES、GMM等4种聚类算法在UCI Online Retail零售商数据集上的客户分类效果;并利用轮廓系数、卡林斯基-哈拉巴斯指数(CHI)和戴维森堡丁指数(DBI)评价比较上述4种聚类算法的分类结果。实证结果表明:在所选零售商数据集上,K-means和AGNES算法的聚类效果较好,DBSCAN和GMM算法的聚类效果不理想,旨在为机器学习聚类算法在基于RFM模型的客户分类提供参考和借鉴。建议企业重视产出数据,完善企业数据相关制度;结合客户数据特征和企业自身销售特点,有针对性地使用聚类算法进行客户细分,辅助总结客户画像,进而制定有针对性的营销策略。

Customer relationship management as an important part of enterprise management,its customer segmentation function directly affects the enterprise marketing strategy.In order to better segment the retail market customers,the RFM model and four clustering algorithms including K-means,DBSCAN,AGNES and GMM are verified in UCI Online Retail by applying a British retailer data set.The results of customer classification on retail retailer data set are compared with those of the above four clustering algorithms by using profile coefficient,Kalinsky Harabas Index(CHI) and Davidson Burger Index(DBI).The empirical results show that K-means and AGNES algorithm have better clustering effect on the selected retailer data set,while DBSCAN and GMM algorithm have less clustering effect,aiming to provide reference and reference for Machine Learning Clustering Algorithm in customer classification based on RFM model.It is recommended that enterprises attach importance to the output data,improving the system related to enterprise data,and combining the characteristics of customer data with the sales characteristics of the enterprise itself to use clustering algorithms for targeted customer segmentation,assisting in summarizing customer profiles,and thus developing targeted marketing strategies.