Spatiotemporal Dynamic Network Modeling for Big Data Analysis of Research Collaboration

As scientific research collaboration becomes increasingly complex and dynamic, traditional static network analysis struggles to capture its spatiotemporal evolution. This paper presents a big data analysis framework for research collaboration based on spatiotemporal dynamic networks, integrating Spatio-Temporal Graph Convolutional Networks (ST-GCN) and multimodal data fusion techniques. The framework uncovers dynamic evolution patterns in collaboration networks and accurately predicts potential collaborations. By constructing weighted co-authorship, keyword, and institutional networks, this study analyzes high-performing computer science scholars, revealing "small-world" characteristics and core-periphery evolution paths. Experimental results show that the ST-GCN model outperforms traditional methods in collaboration prediction, achieving an AUC of 0.9708 and an F1 score of 0.9553. This research offers a novel methodological approach for dynamic collaboration analysis and provides practical insights for research management and policy-making.