Explaining project success through Support Vector Machine approaches: an explainable lens of operations readiness in infrastructure projects

Infrastructure projects exhibit high-dimensional interactions, nonlinearity, and dynamic lifecycle evolution that conventional forecasting methods struggle to capture, thus reducing the predictive accuracy and robustness required for optimal decision-making. In response, this study develops and evaluates an explainable machine learning algorithm based on Support Vector Machine using data obtained from post hoc analysis of 633 operations, project and Operational Readiness and Transition (ORAT) professionals. Readiness factors (‘Facilities,’ ‘People’, ‘Technology’, and ‘Organisation’), were assessed against ‘Project success’, grouped into stronger- and weaker-performing categories. A nonlinear predictive model with adjusted class weighting was trained and tested, demonstrating acceptable overall discrimination, with high sensitivity to higher-success cases but weaker specificity for lower-success cases. Interpretability analysis consistently identified ‘Organisation readiness’ as the most influential factor, with model-derived indicative thresholds emerging for each factor as approximate reference points. Findings show that pairing nonlinear modelling with explainable analytics offers managers a transparent, evidence-based tool for guiding decisions in complex infrastructure project environments.