A hybrid machine learning-enhanced MCDM model for transport safety engineering

Abstract

Abstract Delivering reliable decision recommendations and policy inferences is essential for multi-criteria decision-making (MCDM) processes, particularly for transport safety engineering. This study proposes a hybrid machine learning-enhanced MCDM model that integrates distance correlation-based criteria importance through intercriteria correlation (DCRITIC), weighted aggregated sum product assessment (WASPAS), and K-means clustering, referred to as the DCRITIC–WASPAS–K-means model. In particular, we incorporated a machine learning tool (i.e., a graph-based technique) into the model to effectively and robustly select initial centroids. This integration addresses the uncertainty in traditional k-means clustering, which arises from varying initial centroids and its sensitivity to outliers, especially in datasets with noisy or skewed data points, and, more importantly, reduces the number of iterations and runtime cost. This approach improves the robustness and reliability of decision outcomes, thereby supporting more credible and actionable policy interventions. A case study involving transport safety engineering in the Organization of American States (OAS) region validates the model’s practical utility. Comparative analyses demonstrate its superior performance in ensuring consistent decision outputs and communicating policy implications effectively. The proposed framework provides public administrators, policymakers, and government agencies with a reliable, scalable, and data-driven tool for strategic planning and resource allocation in uncertain environments.