基于多模型融合的电子商务客户流失预测模型研究  

Prediction model research of e-commerce customer churn based on multi-model fusion

作　　者：马文斌 陈硕峰 Ma Wenbin;Chen Shuofeng(Department of Academic Affairs,Guangxi University of Finance and Economics,Nanning 530007,China)

机构地区：[1]广西财经学院教务处,广西南宁530007

出　　处：《无线互联科技》2022年第17期143-145,共3页Wireless Internet Technology

基　　金：2019年度广西高校中青年教师科研基础能力提升项目,项目编号:2019KY0661;2022年应用经济学广西一流学科建设项目资助,项目编号:2022GSXKB05。

摘　　要：由于电子商务客户的非契约性,其流失率较高。为提高客户留存率,降低客户流失率,提高企业核心竞争力,文章提出一种基于多模型融合的电子商务客户流失预测模型。该模型首先以泛化性能较好,且相互存在差异性的随机森林、XGBoost、LightGBM为基分类器,然后以投票法为集成策略,融合多个基分类器的输出作为最终的预测结果。实验结果表明,与随机森林等基模型相比,该方法在F1、Recall、Precision上均有提升,有助于降低客户流失率,提高企业的经济效益。Due to the non-contractual nature of e-commerce customers,their churn rate is high.In order to improve customer retention rate,reduce customer churn rate and improve the core competitiveness of enterprises,a prediction model of e-commerce customer churn based on multi-model fusion is proposed.The model firstly takes random forest,XGBoost and LightGBM,which have good generalization performance and are different from each other,as the base classifier,and then uses voting method as the integration strategy to fuse the output of multiple base classifiers as the final prediction result.Experimental results show that compared with random forest and other base models,this method improves in F1,Recall and Precision,which helps to reduce customer turnover rate and improve economic benefits of enterprises.

关 键 词：电子商务 非契约 客户流失预测 模型融合 投票法 

分 类 号：F713.36[经济管理—产业经济] TP181[自动化与计算机技术—控制理论与控制工程]