Supply Chain Data Hub Logo

Link Prediction

Yunbo Long 507 words 3 minutes Graph Neural Networks Network Analysis Supply Chain Visibility

Supply chain networks are inherently complex, with thousands of firms connected through buyer-supplier relationships that are often opaque and difficult to map. Link prediction, the task of inferring missing or future connections in a network, has emerged as a powerful tool for improving supply chain visibility. By leveraging the structural properties of known supply chain graphs, researchers can predict previously unknown trading relationships, uncover hidden dependencies, and anticipate how networks may evolve over time. This capability is critical for risk management, as disruptions frequently propagate through links that firms did not know existed.

Traditional approaches to supply chain link prediction relied on network topology features such as common neighbours, Jaccard coefficients, and preferential attachment scores. While effective as baselines, these heuristic methods struggle to capture the rich, non-linear patterns present in real-world supply chain graphs. The advent of graph neural networks (GNNs) has significantly advanced the field, enabling models to learn node and edge representations that encode both local neighbourhood structure and global network context. Methods based on graph convolutional networks, graph attention networks, and variational graph autoencoders have demonstrated strong performance on automotive, electronics, and other industry supply chain datasets.

Recent work has further pushed the boundaries by incorporating temporal dynamics, node attributes such as industry classification and geographic location, and multi-relational graph structures. Researchers have also explored how link prediction can support supply chain due diligence, sustainability auditing, and resilience planning. The publications listed below represent key contributions to this rapidly growing area of research, spanning foundational network science approaches through to state-of-the-art deep learning methods.

List of Publications

  1. Brintrup, A., Ledwoch, A. and Barber, J., 2018. Connected cars: Automotive supply chain networks mapped. Cambridge Service Alliance Report. [PDF]

  2. Kosasih, E.E., Brintrup, A., Primber, N. and Kollias, S., 2021. On the use of graph neural networks for supply chain link prediction. IFAC-PapersOnLine, 54(1), pp.1122-1127. [PDF]

  3. Kosasih, E.E. and Brintrup, A., 2022. A machine learning approach for predicting hidden links in supply chain with graph neural networks. International Journal of Production Research, 60(17), pp.5380-5393. [PDF]

  4. Aziz, A., Kosasih, E.E., Huang, R., Thanapalasingam, T. and Brintrup, A., 2023. Link prediction on complex networks: An experimental survey. Data Science and Engineering, 8(3), pp.253-278. [PDF]

  5. Xu, L., Kosasih, E.E. and Brintrup, A., 2023. A temporal graph neural network for supply chain link prediction. Proceedings of the AAAI 2023 Workshop on AI for Supply Chain (AI4SC). [PDF]

  6. Xu, L. and Brintrup, A., 2024. Multi-relational graph neural network for supply chain link prediction. International Journal of Production Research, 62(18), pp.6628-6645. [PDF]

  7. Wichmann, P., Brintrup, A., Baker, S., Woodall, P. and McFarlane, D., 2020. Extracting supply chain maps from news articles using deep neural networks. International Journal of Production Research, 58(17), pp.5320-5336. [PDF]