Feature extraction for artificial intelligence enabled food supply chain failure mode prediction

Abstract

The Farm to Fork Strategy of the European Commission is a contingency plan aimed at always ensuring a sufficient and varied supply of safe, nutritious, affordable, and sustainable food to citizens. The learning from previous crises such as COVID-19 indicates that proactive strategies need to span numerous levels both within and external to food networks, requiring both vertical and horizontal collaborations. However, there is a lack of systematic performance management techniques for ripple effects in food supply chains that would enable the prediction of failure modes. Supervised learning algorithms are commonly used for prediction (classification) problems, but machine learning struggles with large data sets and complex phenomena. Consequently, this research proposes a manual approach to feature extraction for artificial intelligence with the aim of reducing dimensionality for more efficient algorithm performance, and improved interpretability/explainability for benefits in terms of ethics and managerial decision-making. The approach is based on qualitative comparative analysis informed by in-depth case knowledge which is refined through Boolean logic, yielding solutions that reflect complex causality as opposed to single failure point modes. Two case exemplars are presented to support the proposed framework for implementation: export readiness of dairy supply chains under the Russia-Ukraine war, and egg supply chain sustainability during COVID-19 lockdown in the United Kingdom.