基于机器学习的商品特性表征和客户偏好预测  

PRODUCT FEATURES CHARACTERIZATION AND CUSTOMERS’PREFERENCES PREDICTION BASED ON MACHINE LEARNING

作　　者：王辉[1] 李昌刚[1] Wang Hui;Li Changgang(Zhejiang Wanli College,Ningbo 315100,Zhejiang,China)

机构地区：[1]浙江万里学院,浙江宁波315100

出　　处：《计算机应用与软件》2022年第9期158-166,共9页Computer Applications and Software

摘　　要：为分析预测电商企业客户的购买偏好及其对应的商品特性表征,结合K-means与PCA来生成相对应且可解释的商品与客户集群,并将原始数据转换为可用于偏好预测分类器训练的数据。在此基础之上,进一步提出使用Stacking集成学习方法,提高预测准确性与模型泛化能力。研究以某公司的真实客户购买数据来验证所提方法的有效性,实验结果表明,基于购买数据正确地生成了5类商品聚类簇与11类客户聚类簇,并且提高了预测精度,以81.37%的准确率预测出了客户的偏好类别。To predict electricity enterprise customers’buying preferences and analyze the corresponding product feature representation,the K-means and PCA algorithm were integrated to generate the corresponding explainable products and customers clusters.And the raw data were converted to the data that could be used in perference prediction classifier training.On this basis,the Stacking integration method was used to improve the prediction accuracy and model generalization ability.Real purchase data of a certain company was studied to verify the effectiveness of the proposed method.The results show that the product and customer cluster are correctly generated based on purchase data with number 5 and 11.This method improves the prediction accuracy,and predicts the customer preference category with accuracy of 81.37%.

关 键 词：客户偏好预测 商品聚类 K-MEANS PCA STACKING 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]