A decision making framework for traffic noise prediction using artificial intelligence techniques

Road traffic is one of the primary contributors to environmental noise pollution. The study presents a literature review on traffic noise exposure, prediction models, comparative performance and explainability of prediction models, traffic noise action plans, and a conceptual decision making framework in urban environments. In total, 136 studies from 1999 to 2026 were reviewed and synthesized. A Boolean search approach was used to understand the different aspects of traffic noise prediction. The results from field studies showed that conventional models achieved R 2 values ranging from 0.54 to 0.96, whereas machine learning (ML) models achieved R 2 values ranging from 0.81 to 0.99, although these are indicative values from heterogeneous studies and not a strict ranking. Moreover, deep learning (DL) in some studies has demonstrated stronger performance on large datasets. The present study also analyzed the interpretability constraints of ML and DL models, which are further addressed by the utilization of explainable artificial intelligence (AI) algorithms. The key research gap is that very few studies employed cross-study validation of the developed noise models, and there remains a shift needed from prediction oriented modelling to decision oriented noise management. To address these gaps, the study proposes a conceptual decision making framework coupled with a traffic noise action plan to offer a structured pathway for sustainable environmental management.