RMNW: Research leadership recommendation in research leading- participating multiplex networks based on Wasserstein Distance

Effective research leadership recommendation is crucial for forming impactful teams, yet existing methods overlook research leadership dynamics and inadequately measure interlayer similarity of multiplex network layers. In this paper, we propose \(\:RMNW\), a novel model for recommending research leadership relationships (leading-leading, leading-participating) using a two-layer multiplex network: a target layer (research leadership) and an auxiliary layer (participating-participating). The \(\:RMNW\) quantifies interlayer similarity via Wasserstein Distance applied to the distributions of nodes’ local and global neighborhoods, capturing structural alignment beyond simple centrality correlations. It integrates target and auxiliary layer information using a tunable parameter (\(\:\lambda\:\)). Evaluated on 426708 publications in Pharmaceutical Sciences (30286 authors), 52381 publications in Information Science & Library Science (6853 authors), and 173264 publications in Computer Sciences (18382 authors). The \(\:RMNW\) significantly outperforms state-of-the-art baselines. Key results include: 2.42% higher \(\:F1@7\), 4.63% higher \(\:MRR\) and 4.7% higher \(\:nDCG\:\)than the best baseline (\(\:MALM\)) in Pharmaceutical Sciences. Optimal performance at \(\:\lambda\:\)=0.5, demonstrating the critical role of auxiliary layer information (performance drops 36–46% without it). Superiority over variants using degree-correlation (\(\:RMN{W}_{DDC}\)), neighbor-overlap (\(\:RMN{W}_{ASSN}\)), or link-overlap (\(\:RMN{W}_{LO}\)), confirming Wasserstein Distance’s effectiveness. The \(\:RMNW\) enables precise leader/participant selection for researchers, enhances institutional network growth, and informs funding allocation.