Scholarlink

<p>Supply chain disruption prediction faces significant challenges due to the system’s inherent complexity and dynamic uncertainty. While existing graph learning approaches utilize knowledge graphs (KGs) to capture semantic links, they predominantly rely on pairwise connections, failing to model the high-order group interactions critical for systemic risk identification. To bridge this gap, we construct a comprehensive supply chain risk knowledge graph and propose a novel hypergraph dynamic graph attention neural network (HG-DRA). Distinct from traditional methods, HG-DRA introduces two core innovations: (1) hypergraph representation learning, which explicitly models complex <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(N\)</EquationSource> <EquationSource Format="MATHML"><math> <mi>N</mi> </math></EquationSource> </InlineEquation>-ary correlations among enterprises to capture macro-level structural risks; and (2) a hierarchical dynamic relational attention mechanism, which adaptively aggregates heterogeneous operational features by filtering noise and prioritizing risk-relevant relations. However, modeling these high-order correlations in massive networks involves computationally intensive matrix operations. To address the scalability and real-time requirements, we leverage high-performance computing (HPC) techniques to accelerate the training and inference of the proposed model. Experiments show that HG-DRA can effectively integrate multi-dimensional features and complex topological structures.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Supply chain disruption risk prediction based on hypergraph representation and dynamic relational-attentive

  • 期刊论文

  • 发表日期：

    2026年2月26日

查看全文
  • Jinlong Wang,
  • Qixin Zhao,
  • Yingmin Liu,
  • Pengjun Li,
  • Yuanyuan Zhang,
  • Xiaoyun Xiong

摘要

Supply chain disruption prediction faces significant challenges due to the system’s inherent complexity and dynamic uncertainty. While existing graph learning approaches utilize knowledge graphs (KGs) to capture semantic links, they predominantly rely on pairwise connections, failing to model the high-order group interactions critical for systemic risk identification. To bridge this gap, we construct a comprehensive supply chain risk knowledge graph and propose a novel hypergraph dynamic graph attention neural network (HG-DRA). Distinct from traditional methods, HG-DRA introduces two core innovations: (1) hypergraph representation learning, which explicitly models complex \(N\) N -ary correlations among enterprises to capture macro-level structural risks; and (2) a hierarchical dynamic relational attention mechanism, which adaptively aggregates heterogeneous operational features by filtering noise and prioritizing risk-relevant relations. However, modeling these high-order correlations in massive networks involves computationally intensive matrix operations. To address the scalability and real-time requirements, we leverage high-performance computing (HPC) techniques to accelerate the training and inference of the proposed model. Experiments show that HG-DRA can effectively integrate multi-dimensional features and complex topological structures.