基于Pareto/NBD的用户价值预测模型研究  

Prediction Model of Consumer Value Based on Pareto/NBD

作　　者：李征仁 张晓航[1] 王光辉 石东海 石文华[1] LI Zheng-ren;ZHANG Xiao-hang;WANG Guang-hui;SHI Dong-hai;SHI Wen-hua(School of Economics and Management, Beijing University of Posts and Telecommunications, Beijing 100876, China;Institutes of Science and Development, Chinese Academy of Sciences, Beijing 100190, China;Didi (China) Technology Co., Ltd, Beijing 100193, China)

机构地区：[1]北京邮电大学经济管理学院,北京100876 [2]中国科学院科技战略咨询研究院,北京100190 [3]滴滴(中国)科技有限公司,北京100193

出　　处：《北京邮电大学学报（社会科学版）》2020年第3期7-14,共8页Journal of Beijing University of Posts and Telecommunications(Social Sciences Edition)

基　　金：国家自然科学基金资助项目(71974182)。

摘　　要：为提高用户价值预测的准确率,在已有Pareto/NBD模型基础上,结合有监督的机器学习算法提出改进模型,即使用最近多个子时间段的购买行为数据来共同预测未来购买频次,充分利用用户历史消费信息。最后,使用滴滴(中国)科技有限公司用户消费数据对改进的模型进行验证。结果表明,改进后的Pareto/NBD模型显著提高了预测准确率。To improve the prediction accuracy of consumer value,based on the existing Pareto/NBD model and the supervised machine learning algorithm,an improved model is proposed,that is,the purchasing behavior data of recent sub-periods are used to predict the future purchasing frequency,which makes full use of users'historical consumption information.The improved model was verified by using DiDi's user consumption data,and the results show that the proposed model can improve the prediction accuracy significantly.

关 键 词：用户生命周期价值 Pareto/NBD 有监督学习 

分 类 号：F713.50[经济管理—市场营销] F224[经济管理—产业经济]