关键词:  , LDA主题模型, 双模网络, 图片特征, 销售额分析, 口碑营销

Abstract: With the development of internet technology, electronic word - of - mouth, consumers’ consumption habits and merchants’ online publicity have all brought important influences on the turnovers of merchants.  The purpose of this paper is to explore the roles of online review text data, network data between merchants and consumers, publicity picture data of merchant and other diversified information in influencing merchants' sales.  The research shows that the topic factors of comment text, the characteristics of merchant centrality and user centrality of two-mode network, and the publicity pictures' degree of balance all have significant effect on merchants' sales.  In terms of the topics of the comment content, two comment topics "volume" and " overall" have significantly positive impacts on the sales of the merchants. When neighboring merchants have great influences in the network, they will significantly reduce the sales of the merchants.  Thus, merchants need to differentiate user groups to expand their own unique users. In addition, it is necessary to attract active and influential user groups to increase the sales of merchants.  In the aspect of picture selection, merchants are recommended to provide pictures with high degree of color balance and pay attention to the color collocation of the picture's foreground and background.  Symmetric pictures make users feel comfortable and can boost the sales of merchants.  Finally, the paper subdivides the merchants according to the model  results and analyzes the advantages and disadvantages of each type of the merchants in different characteristics.  Suggestion is made on boosting offline sales according to different clusters of merchants.

Key words: LDA topic model ,  , two-mode network ,  , picture features ,  , sales analysis ,  , word-of-mouth marketing