A fusion method for large-scale online review ranking

Abstract

Online reviews constitute a pivotal informational asset that significantly influences consumer purchasing decisions. While prior research has predominantly focused on product ranking based on these reviews, the challenge of handling large-scale online review datasets has been largely overlooked. Furthermore, the consensus among ranking outcomes, a critical factor in ensuring the reliability of rankings, has seldom been addressed. This study introduces a novel consensus-based ranking approach tailored for large-scale rating datasets, incorporating cluster analysis, multi-attribute decision-making (MADM), and a consensus-reaching mechanism. Initially, a rating matrix is constructed to consolidate the extensive rating data. Subsequently, cluster analysis segments the vast user base, and MADM is leveraged to produce group-specific rankings. Ranking aggregation technique is then applied to synthesize the rankings from disparate user groups into a unified collective ranking. Ultimately, a consensus-reaching process, which accounts for both intra-cluster and inter-cluster agreement, refines the ranking to ensure a harmonized consensus among all user groups. The efficacy and applicability of this methodology are substantiated through an empirical case study examining hotel rankings in New York City.