Answers are grounded in this paper and cite their sources.
Study models signal propagation uncertainty in complex networks, suggesting a novel approach for ESG risk governance.
With the development of green finance and the complexity of the social network ecosystem, ESG risk governance and social network risk prevention and control have become core issues in the intelligent management of complex networks. Both belong to typical complex heterogeneous networks and share common topological characteristics such as multi-agent interaction and uncertainty of signal propagation. However, their inherent heterogeneity and non-linear geometric structure make it difficult for traditional models to collaboratively capture the semantic correlation and geometric features of networks, which seriously restricts the accuracy of risk analysis. To this end, this paper proposes the Riemannian Geometry-Aware Heterogeneous Graph Network (RGA-HGN), which deeply integrates Riemannian geometry and heterogeneous graph attention mechanisms to realize the unified modeling of signal propagation uncertainty in social networks and ESG governance networks, as well as the accurate prediction of cross-agent risks. The model unifies multi-type node features through type-aware Euclidean embedding, retains the inherent geometric structure of the network by virtue of Riemannian manifold projection, and quantifies network risk correlation and captures the uncertainty of signal propagation using Riemannian geometry attention. Based on multi-source public data, a heterogeneous network of 30 Dow Jones enterprises is constructed, and the model is compared with 9 benchmark models on three core tasks. The results show that RGA-HGN significantly outperforms all baseline models. This study fills the research gap of geometric deep learning in the fusion analysis of social and ESG heterogeneous networks, and provides a universal and generalizable framework for complex network risk analysis.
No takes yet. Share an insight, caveat, or question.
Jixian Zhang (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: