AI-Driven Statistical Modeling for Social Network Analysis

In this paper, we present a novel AI-driven statistical modeling approach for social network analysis, specifically focusing on information dissemination forecasting. The proposed Paradigm-Conscious Bistream Dissemination Framework (PCBDF) addresses the limitations of existing models by simultaneously capturing dynamic customer preferences and chain dependencies. Our framework leverages bistream graph neural networks to separately learn customer and chain embeddings. Specifically, the dynamic graph convolutional network captures customer repost preferences at various granularities, while the hyper-graph convolutional network learns chain hypergraphs and customer relationships. By integrating paradigms of chain embeddings, we enhance the accuracy of information dissemination forecasting. Quantitative evaluations on multiple datasets demonstrate that PCBDF achieves significant improvements over state-of-the-art models, with an increase in mean average precision by up to 14.43% and in Hits scores by up to 16.11% across multiple datasets.