Supply chain risk management and artificial intelligence: state of the art and future research directions

Abstract

Supply chain risk management (SCRM) encompasses a wide variety of strategies aiming to identify, assess, mitigate and monitor unexpected events or conditions which might have an impact, mostly adverse, on any part of a supply chain. SCRM strategies often depend on rapid and adaptive decision-making based on potentially large, multidimensional data sources. These characteristics make SCRM a suitable application area for artificial intelligence (AI) techniques. The aim of this paper is to provide a comprehensive review of supply chain literature that addresses problems relevant to SCRM using approaches that fall within the AI spectrum. To that end, an investigation is conducted on the various definitions and classifications of supply chain risk and related notions such as uncertainty. Then, a mapping study is performed to categorise existing literature according to the AI methodology used, ranging from mathematical programming to Machine Learning and Big Data Analytics, and the specific SCRM task they address (identification, assessment or response). Finally, a comprehensive analysis of each category is provided to identify missing aspects and unexplored areas and propose directions for future research at the confluence of SCRM and AI.

Keywords:

• supply chain risk management
• artificial intelligence
• decision-making
• SCRM strategy
• supply chain disruption

Disclosure statement

No potential conflict of interest was reported by the authors.

Supplemental data

Supplemental data for this article can be accessed at https://doi.org/10.1080/00207543.2018.1530476.

ORCID

George Baryannis http://orcid.org/0000-0002-2118-5812

Sahar Validi http://orcid.org/0000-0003-1042-1856

Grigoris Antoniou http://orcid.org/0000-0003-3673-6602

Notes

1. https://www.ibm.com/watson/

2. https://deepmind.com/research/alphago/

3. For a detailed analysis of relevant literature, please refer to Section 3.3

4. http://www.elsevier.com/solutions/scopus/content

5. https://clarivate.libguides.com/webofscienceplatform/coverage

6. The survey dataset can be provided in various formats (e.g. SciVal, EndNote, or .BIB/.RIS files) upon request to the corresponding author.

7. Note that probabilistic constraints are included in other studies as well, but as part of a hybrid approach, hence they are discussed in Section 5.4.

8. Bayesian networks are used in these studies only for inference. Studies that use such models for learning as well are classified within the Machine Learning and Big Data category, discussed in Section 5.8