Multisource Data Driven Product Ranking Model With Heterogeneous Customers

In order to improve the quality of consumer decision making, ranking products is an important task for e-commerce platforms. Yet, prior product ranking methods usually consider one kind of product information (that is, online text reviews), and fail to systematically and comprehensively consider the heterogeneity of consumers with individual's idiosyncratic taste and their multidimensional product preferences. Based on which, this article proposed a multisource data-driven product ranking model in which the consumers could interact online to reflect their personalized preferences and get a personalized product ranking list with less effort. First, multisource evaluative information (e.g., product parameters, online ratings, and review text) was collected by ecommerce platforms, and then the fusing data was obtained by sentiment analysis and two-step fusion process. Second, according to the consumer's familiarity with products, three corresponding rules for determining attribute weight were proposed and a utility function was constructed by uniting absolute utility of the product itself and the relative utility comparing the reference point. Finally, a case of purchasing new energy vehicles was applied to illustrate the practicability of the proposed model. The results showed that the proposed ranking model can obtain the personalized product list with heterogeneous consumers.