Institutional Sentiment as an Early Warning System for Supply Chain Risk: A Machine Learning Approach

Supply chain risk has intensified amid growing global uncertainty, creating the need for advanced analytical approaches that provide timely early warning signals for operational decision-making. Traditional risk models largely rely on market indicators and historical performance data, while institutional communications that shape global economic and commodity conditions remain underutilized. This study develops a machine learning–based framework for supply chain risk modeling by extracting sentiment and thematic signals from large-scale institutional textual information. Using freely available secondary datasets, the research applies sentiment analysis, topic modeling, and predictive learning techniques to transform unstructured text into quantitative risk indicators. These indicators are linked to operational risk outcomes such as commodity price volatility, disruption frequency, and lead time variability through time-series forecasting models. The results show that institutional sentiment significantly enhances the prediction accuracy of supply chain risk events beyond conventional economic measures, offering effective early warning capabilities for operations planning and procurement strategies. The study contributes to operations and supply chain management literature by introducing institutional text analytics as a novel and scalable source of risk intelligence. From a managerial perspective, the findings demonstrate how data-driven risk forecasting can support proactive mitigation, improved resilience, and more robust supply chain design in volatile environments.