Engineering Applications of Artificial Intelligence, vol.163, 2026 (SCI-Expanded, Scopus)
Supply Chain Digital Transformation Projects (SCDTPs) are crucial for boosting the resilience of manufacturing supply chains. Amid complex and volatile risks, sound risk response decisions (RRDs) are imperative. However, the extended project duration and advanced digital technologies lead to limited expert assessments, thus disturbing decisions. It is an aspect overlooked in existing models. To address this gap, this study develops an online learning-driven fuzzy dynamic RRD model tailored to SCDTPs. First, Online Learning Z-numbers (OLZs) are proposed by integrating online learning algorithm and Z-numbers, enabling continuous updating of assessments as projects evolve. Second, a hybrid method combining the FUll COnsistency Method (FUCOM) and Gravity Model (GM), known as FUCOM-GM, is employed to prioritize risks. Third, within the OLZs framework, a dynamic optimization model maximizing total utility is formulated to determine the optimal RRD at each project stage. The model developed in this study is validated via an illustrative example. Sensitivity and comparative analyses show that 1) Excessive risk aversion or preference undermines budget utilization efficiency; 2) RRDs that account for differences across project stages are more effective; 3) Online learning algorithm can enhance decision capabilities. This study offers theoretical and methodological insights for SCDTP risk response, contributing to the enhancement of manufacturing supply chain resilience.