Identifying the Value Identity Structure of Youth in the Context of Red Culture Dissemination Using GNN

Abstract

Understanding information propagation mechanisms in large-scale heterogeneous networks requires effective representation of complex relational structures and diffusion dynamics. This study proposes a graph neural network (GNN)-based framework for topology-aware identification of structural communities and propagation patterns in cross-platform information dissemination systems. A heterogeneous information graph is constructed by integrating user interactions, content transmission behaviors, and platform-level communication relationships. To capture high-order structural dependencies, a relation-aware message-passing mechanism is developed to jointly encode network topology and semantic correlations into unified node representations. Based on the learned embeddings, structural communities, influential nodes, and dominant propagation paths are quantitatively identified to characterize information-flow patterns within the network. Experimental results demonstrate that the proposed framework detects 19 core structural communities with an average degree centrality of 0.71 and an internal edge density of 0.62, outperforming conventional network-analysis approaches in structural representation capability. The proposed framework provides an effective methodology for heterogeneous network modeling, information propagation analysis, influence-path characterization, and graph-based representation learning, offering potential applications in intelligent communication systems, distributed information networks, and complex propagation-environment analysis.