基于隐含狄利克雷分配模型的消费者在线评论复杂网络构建及其应用

摘要

为研究消费者在线评论的相互关系及整体演化发展,以隐含狄利克雷分配模型对消费者在线评论进行话题挖掘为基础,通过Pearson相似度确定评论间话题关系,构建了以评论为节点的复杂网络模型.以京东等平台上的手机评论语料库为例,对相应复杂网络的拓扑特性进行分析.研究表明,相同产品相同规模的评论网络统计特性相近;而相同平台上同一产品不同规模的评论网络性质差异明显,即小规模评论(产品销售初期)构成的网络是一个无标度网络,大规模评论(产品销售中后期)构成的网络演化为一个度分布为钟形分布的小世界网络.揭示了消费者发表在线评论行为之间相互作用,会影响整体评论网络的拓扑性质和演化规律,并为市场营销和管理实践提出了指导性建议.%In order to study the relationship and the evolution of the consumer online reviews,this paper proposes a complex network model with reviews as nodes by calculating reviews' topics with latent Dirichlet allocation model and topic similarities among reviews with Pearson similarity.Taking the mobile phone reviews as an example,analysis is undertaken and the results indicate that the statistical characteristics of the review network with the same scale of the same product are similar;as for different scales,the characteristics vary significantly.Small-scale review(the initial stage of product sales) network is a scale-free network.Large-scale review (the middle and late stage of product sales) network evolves into a small world network with bell shaped degree distribution.It reveals the interaction of online review behaviors among consumers which influences the topological properties and the evolution rules of the review network.Instructional suggestions are put forward for the marketing and management practices.