Graph Neural Networks and Network Science for Author Impact, Citation Modeling, and Scholarly Analytics

Abstract

Graph neural networks and network science form a powerful foundation for modeling scholarly ecosystems, where authors, papers, venues, and citations interact within complex structures. Citation graphs and coauthorship networks encode rich relational patterns that influence scientific influence, topic diffusion, and the evolution of knowledge domains. This article investigates the integration of graph neural architectures with network science principles to improve author impact assessment and citation prediction. The study develops a unified framework for analyzing structural properties of scholarly graphs, learning expressive node embeddings, forecasting citation trajectories, and deriving community level indicators. Empirical evaluation demonstrates that graph based learning enhances predictive accuracy and offers deeper insight into the dynamics of scholarly communication.