Mapping Scientific Frontiers: Network Embeddings Reveal Hidden Structures in Global Research Mobility

Abstract

Understanding global scientific mobility requires a framework that captures the complex relationships between research institutions beyond geographic proximity. This study employs network-based embedding techniques to model co-affiliation trajectories of mobile scientists as a dynamic, high-dimensional network. By leveraging word2vec-inspired representations, we uncover latent connectivity structures that encode institutional prestige, regional clustering, and linguistic affiliations. Our approach demonstrates that functional distances derived from embeddings outperform traditional geographic models in predicting mobility patterns. Furthermore, hierarchical clustering of embedded institutions reveals an implicit global research network shaped by cultural and economic factors. These findings highlight the power of network embeddings in deciphering the intricate web of global scientific exchange.