Platforms empower: Mining online reviews for supporting consumers decisions

Abstract

With the progress of information technology, various platforms have emerged and rapidly developed. In product recommendation platforms, online reviews generated by consumers, as a key source of information, exert a substantial influence on purchasing decisions made by consumers. Although prior research has made some progress in this field, there is still a lack of exploration on the types of reviews information, the sentiment tendencies, and consumer decision-making behavior. Guided by text mining techniques and behavioral decision theory, this paper develops a heterogeneous data-driven decision-support model to more comprehensively extract information from online reviews and gain insights into consumer purchasing behavior. To handle the heterogeneity of online reviews, sentiment analysis is conducted to convert unstructured text data into sentiment values with structurization. Thereafter, a three-stage heterogeneous data aggregation framework is developed to define overall evaluation by fusing unstructured text reviews and structured star ratings. After defining a new attribute called word-of-mouth effect (WoME) based on interactive behavior data (such as views, likes and replies), we present a product ranking method by integrating regret theory and the logarithmic TODIM (LogTODIM) method. Furthermore, a case study is presented that evaluates the ranking of new energy vehicles (NEVs) on the Autohome platform, thereby verifying the feasibility of the proposed model.