Integrated Collaborative Scholar Network Prediction Framework (ICSNPF): A Hybrid Approach Analysing Content, Structural and Temporal Attributes

This study presents an advanced approach to link prediction in academic co-authorship networks, leveraging graph-based, content-based, and temporal features to enhance predictive accuracy. Conventional link prediction methods ignore the dynamic nature of research collaborations and subject dynamics in favour of static structural features. To overcome this constraint, we suggest a Graph-Content- Temporal (GCT) link prediction model that combines Temporal Graph Networks (TGAT) to capture temporal dependencies in co-authorship evolution and Graph Neural Networks (GNNs) for structural feature ex- traction. Additionally, an attention-based fusion mechanism is used in the study to dynamically weigh the contributions of various features. With an F1-score of 0.8527 and an AUC of 0.8320, the suggested model clearly surpasses state-of-the-art techniques, according to evaluation on the APS Metadata and Citation Pairs Dataset (2022). These results underscore the importance of multi-perspective feature fusion in academic network analysis and set a new standard for forecasting scholarly interactions.