摘要
文章基于消费者点击流数据和网络结构,使用时态指数随机图模型(TERGM)和消费者点击流数据建构了消费者动态共同购买网络,从产品点击次数、相对浏览时间、好评数、差评数和产品入度等维度测度了影响消费者共同购买行为发生的关键变量,并与指数随机图模型(ERGM)进行了比较。结果表明:产品相对浏览时间、好评数和产品入度促进消费者共同购买行为发生,而产品点击次数会降低消费者共同购买可能性;TERGM模型适合于消费者共同购买行为的网络分析,且拟合效果优于ERGM,验证了TERGM模型对消费者共同购买行为的适用性;文章提出点击流的隐式反馈中应加入时间网络结构视角研究对共同购买网络形成的影响,可为推荐系统优化设计提供有益参考。
Based on consumer clickstream data and network structure,this paper uses Temporal Exponential Random Graph Model(TERGM)and consumer clickstream data to construct a consumer dynamic co-purchase network,identifies the key variables of consumer co-purchase behavior from the dimensions of product click times,relative browsing time,positive comments,negative comments and product penetration,and compares it with Exponential Random Graph Model(ERGM).The results show that product relative browsing time,praise number and product penetration promote the occurrence of consumers'co-purchase behavior,while the number of product clicks will reduce the possibility of consumers'co-purchase.TERGM model is suitable for the network analysis of consumers'co-purchase behavior,and the fitting effect is better than ERGM,which verifies the applicability of the TERGM model to consumer co-purchase behavior.This paper suggests that the network structure perspective should be added in the implicit feedback of clickstream to study the impact on the formation of co-purchase network,which provides a useful reference for the optimal design of recommendation system.
作者
易闽琦
温展明
YI Minqi;WEN Zhanming(Guangdong Peizheng College,Guangzhou 510830,China;Guangdong University of Technology,Guangzhou 510520,China)
出处
《现代信息科技》
2024年第12期138-145,共8页
Modern Information Technology
关键词
共同购买
消费者点击流数据
网络分析
指数随机图模型
时间指数随机图模型
co-purchase
consumer clickstream data
network analysis
exponential random graph model
time exponential random graph model
