AN EXPLAINABLE TEMPORAL GRAPH NEURAL NETWORK FOR DISRUPTION PREDICTION AND RISK PROPAGATION IN MULTI-TIER SUPPLY CHAINS
Keywords:
Supply chain disruption, Risk propagation, Ripple effect, Temporal graph neural networks, Explainable artificial intelligence, Attention mechanismAbstract
Global supply chains are increasingly multi-tier, interconnected, and vulnerable to disruptions that do not stay local but cascade upstream-to-downstream through material-flow dependencies—the so-called ripple effect. Effective risk management requires not only predicting which entities will be disrupted, but also explaining why, by tracing a predicted disruption back to its root-cause suppliers. We present X-TGNN, an explainable temporal graph neural network that couples a per-node recurrent temporal encoder with a directed, tier-respecting, edge-weighted multi-head graph-attention mechanism aligned with the direction of risk propagation. The learned attention doubles as an intrinsic, single-forward-pass explanation that attributes each predicted disruption to specific upstream suppliers. Because real multi-tier disruption datasets with ground-truth propagation paths are proprietary and unavailable, we further contribute a high-fidelity simulation testbed whose cascade dynamics are grounded in the supply-chain risk-diffusion literature and which records the true responsible edges for every propagation event—enabling, for the first time in this setting, a quantitative evaluation of explanation faithfulness and correctness. Across a testbed in which 61% of disruptions are propagation-driven, graph models improve new-onset prediction by 2.5× in average precision over non-graph baselines. X-TGNN matches the strongest baseline in prediction while its attention explanations identify true root-cause suppliers with precision@1 of 0.77 and NDCG of 0.97—matching, and in rank correlation exceeding, far more expensive post-hoc gradient and perturbation methods, and vastly better than a random baseline (0.26). The result is a model that is at once accurate, propagation-aware, and inherently explainable.References
[1] Dolgui A, Ivanov D, Sokolov B. Ripple effect in the supply chain: an analysis and recent literature. International Journal of Production Research, 2018, 56(1-2): 414-430.
[2] Ivanov D, Sokolov B, Dolgui A. The ripple effect in supply chains: trade-off ‘efficiency-flexibility-resilience’ in disruption management. International Journal of Production Research, 2014, 52(7): 2154-2172.
[3] Basole R C, Bellamy M A. Supply network structure, visibility, and risk diffusion: A computational approach. Decision Sciences, 2014, 45(4): 753-789.
[4] Ivanov D. Predicting the impacts of epidemic outbreaks on global supply chains: A simulation-based analysis on the coronavirus outbreak (COVID-19/SARS-CoV-2) case. Transportation Research Part E: Logistics and Transportation Review, 2020, 136: 101922.
[5] Kipf T N, Welling M. Semi-supervised classification with graph convolutional networks. In: Proceedings of the International Conference on Learning Representations, 2017.
[6] Veličković P, Cucurull G, Casanova A, et al. Graph attention networks. In: Proceedings of the International Conference on Learning Representations, 2018.
[7] Hamilton W L, Ying R, Leskovec J. Inductive representation learning on large graphs. In: Proceedings of the Conference on Neural Information Processing Systems, 2017.
[8] Xu D, Ruan C, Korpeoglu E, et al. Inductive representation learning on temporal graphs. In: Proceedings of the International Conference on Learning Representations, 2020.
[9] Pareja A, Domeniconi G, Chen J, et al. EvolveGCN: Evolving graph convolutional networks for dynamic graphs. In: Proceedings of the AAAI Conference on Artificial Intelligence, 2020: 5363-5370.
[10] Rossi E, Chamberlain B, Frasca F, et al. Temporal graph networks for deep learning on dynamic graphs. In: ICML Workshop on Graph Representation Learning, 2020.
[11] Ying R, Bourgeois D, You J, et al. GNNExplainer: Generating explanations for graph neural networks. In: Proceedings of the Conference on Neural Information Processing Systems, 2019: 9240-9251.
[12] Luo D, Cheng W, Xu D, et al. Parameterized explainer for graph neural network. In: Proceedings of the Conference on Neural Information Processing Systems, 2020: 19620-19631.
[13] Vu M N, Thai M T. PGM-Explainer: Probabilistic graphical model explanations for graph neural networks. In: Proceedings of the Conference on Neural Information Processing Systems, 2020: 12225-12235.
[14] Pope P E, Kolouri S, Rostami M, et al. Explainability methods for graph convolutional neural networks. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019: 10772-10781.
[15] Hosseini S, Ivanov D, Dolgui A. Review of quantitative methods for supply chain resilience analysis. Transportation Research Part E: Logistics and Transportation Review, 2019, 125: 285-307.
[16] Kleindorfer P R, Saad G H. Managing disruption risks in supply chains. Production and Operations Management, 2005, 14(1): 53-68.
[17] Craighead C W, Blackhurst J, Rungtusanatham M J, et al. The severity of supply chain disruptions: design characteristics and mitigation capabilities. Decision Sciences, 2007, 38(1): 131-156.
[18] Tang C S. Perspectives in supply chain risk management. International Journal of Production Economics, 2006, 103(2): 451-488.
[19] Wagner S M, Bode C. An empirical examination of supply chain performance along several dimensions of risk. Journal of Business Logistics, 2008, 29(1): 307-325.
[20] Inoue H, Todo Y. Firm-level propagation of shocks through supply-chain networks. Nature Sustainability, 2019, 2: 841-847.
[21] Gilmer J, Schoenholz S S, Riley P F, et al. Neural message passing for quantum chemistry. In: Proceedings of the International Conference on Machine Learning, 2017: 1263-1272.
[22] Xu K, Hu W, Leskovec J, et al. How powerful are graph neural networks? In: Proceedings of the International Conference on Learning Representations, 2019.
[23] Seo Y, Defferrard M, Vandergheynst P, et al. Structured sequence modeling with graph convolutional recurrent networks. In: Proceedings of the International Conference on Neural Information Processing, 2018: 362-373.
[24] Li Y, Yu R, Shahabi C, et al. Diffusion convolutional recurrent neural network: Data-driven traffic forecasting. In: Proceedings of the International Conference on Learning Representations, 2018.
[25] Yu B, Yin H, Zhu Z. Spatio-temporal graph convolutional networks: A deep learning framework for traffic forecasting. In: Proceedings of the International Joint Conference on Artificial Intelligence, 2018: 3634-3640.
[26] Sankar A, Wu Y, Gou L, et al. DySAT: Deep neural representation learning on dynamic graphs via self-attention networks. In: Proceedings of the ACM International Conference on Web Search and Data Mining, 2020: 519-527.
[27] Wu Z, Pan S, Chen F, et al. A comprehensive survey on graph neural networks. IEEE Transactions on Neural Networks and Learning Systems, 2021, 32(1): 4-24.
[28] Zhou J, Cui G, Hu S, et al. Graph neural networks: A review of methods and applications. AI Open, 2020, 1: 57-81.
[29] Baldassarre F, Azizpour H. Explainability techniques for graph convolutional networks. arXiv:1905.13686, 2019.
[30] Ribeiro M T, Singh S, Guestrin C. ‘Why should I trust you?’: Explaining the predictions of any classifier. In: Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 2016: 1135-1144.
[31] Lundberg S M, Lee S-I. A unified approach to interpreting model predictions. In: Proceedings of the Conference on Neural Information Processing Systems, 2017: 4765-4774.
[32] Yuan H, Yu H, Gui S, et al. Explainability in graph neural networks: A taxonomic survey. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2022, 45(5): 5782-5799.