Game-Theoretic Approach to Weighted Rank Aggregation Based on Strategic Interactions Among Ranker

Abstract

Rank Aggregation (RA) refers to the process of combining multiple rankings from a set of candidate base rankers to achieve a better overall ranking. It has been widely applied across various domains and plays a critical role in integrating information from different biological studies addressing a common problem. In biological studies, due to the high heterogeneity of sources, the base rankers are often partial, lengthy, and of varying quality, and the ground-truth rankings are typically unavailable. This paper proposes a novel model for RA in biological and biomedical applications, in which the combination of multiple rankings from different sources is modeled as a non-cooperative game among rankers. Unlike traditional methods that treat rankers as independent and passive entities, the proposed approach models each ranker as a strategic agent seeking to maximize its own utility and influence on the final aggregated ranking. The proposed utility function integrates similarity to the final ranking, agreement with other rankers, and the cost of generating a ranking. An iterative algorithm based on best-response dynamics is used to update each ranker's weight according to their current utility, leading to a stable aggregated output. To demonstrate the effectiveness of the proposed method in solving biological problems, three benchmark datasets—Breast, microRNA, and Prostate—were used. Experimental results indicate that the proposed method outperforms traditional  RA techniques in both effectiveness and robustness.