<p>Project disruptions may occur due to unexpected incidents or intentional acts, leading to schedule extensions. The susceptibility of a project to unforeseen occurrences dictates its vulnerability. The assessment of project vulnerability enables managers to incorporate preventive strategies during the planning phase or, in the recovery phase, prioritize projects in decision-making for implementing appropriate strategies. Nonetheless, prior research has mostly neglected the impact of project topology and network structure on performance and susceptibility to interruptions. Additionally, while machine learning has been used in project management, its integration with project network structural metrics to classify projects based on inherent vulnerability is still limited. This study presents an innovative method using network-based metrics to assess project vulnerability. These measurements function as input characteristics for machine learning algorithms to assess vulnerability during the planning phase. A decision tree is first used to derive splitting criteria. A random forest approach, with hyperparameters adjusted using grid search, is used to enhance prediction accuracy. Empirical project data are used to verify the models, which are then evaluated with three alternative algorithms based on established machine learning performance metrics. The assessment findings indicate that the suggested model attains enhanced prediction performance.</p>