Resilience Inference for Supply Chains with Hypergraph Neural Network

TL;DR

This work tackles the problem of inferring supply chain resilience from hypergraph topology and observed inventory trajectories without explicit dynamics. It introduces SCRI and the SC-RIHN model, which integrates a set-based feature encoder, hypergraph convolution, and temporal aggregation to capture higher-order firm–product interactions. Empirical results on synthetic and real-world datasets show SC-RIHN outperforms traditional ML and GNN baselines and remains robust under structural perturbations, illustrating the value of higher-order modeling. The findings highlight the practical potential of hypergraph-based resilience inference for proactive risk management and robust supply chain design.

Abstract

Supply chains are integral to global economic stability, yet disruptions can swiftly propagate through interconnected networks, resulting in substantial economic impacts. Accurate and timely inference of supply chain resilience the capability to maintain core functions during disruptions is crucial for proactive risk mitigation and robust network design. However, existing approaches lack effective mechanisms to infer supply chain resilience without explicit system dynamics and struggle to represent the higher-order, multi-entity dependencies inherent in supply chain networks. These limitations motivate the definition of a novel problem and the development of targeted modeling solutions. To address these challenges, we formalize a novel problem: Supply Chain Resilience Inference (SCRI), defined as predicting supply chain resilience using hypergraph topology and observed inventory trajectories without explicit dynamic equations. To solve this problem, we propose the Supply Chain Resilience Inference Hypergraph Network (SC-RIHN), a novel hypergraph-based model leveraging set-based encoding and hypergraph message passing to capture multi-party firm-product interactions. Comprehensive experiments demonstrate that SC-RIHN significantly outperforms traditional MLP, representative graph neural network variants, and ResInf baselines across synthetic benchmarks, underscoring its potential for practical, early-warning risk assessment in complex supply chain systems.

Figures (5)

• Figure 1: Illustration of supply chain resilience. Two supply chains with different topologies (left) experience the same disruption (a factory fire). The inventory trajectory of a downstream store (right) shows recovery in the upper network (resilient) and persistent failure in the lower network (non-resilient).
• Figure 2: Overview of the proposed Supply Chain Resilience Inference Hypergraph Network (SC-RIHN). At each time step, the Feature Encoder decomposes firm state vectors into individual feature dimensions and combines them with learnable positional embeddings. A multi-layer perceptron (MLP) followed by a pooling operation then projects these heterogeneous inputs from various supply chains into a unified latent space. Subsequently, the Hypergraph Encoder initializes product node features using positional embeddings and integrates them with firm node embeddings. After multiple hypergraph convolution layers capturing higher-order dependencies, the refined firm node embeddings are pooled into a graph-level representation. Finally, the Global Readout layer aggregates these representations across all time steps into a single system-level resilience embedding, which an MLP uses to infer resilience.
• Figure 3: Performance of GNN baselines with different hypergraph reduction strategies on (a) SCR and (b) TESLA.
• Figure 4: Distribution of predicted resilience probabilities on the TESLA test set.
• Figure 5: F1-score (mean over 10 random runs) on SCR under varying window length $T$ and SC-RIHN layer depth $L$.