An automated framework for traffic noise level analysis using explainable artificial intelligence techniques

Traffic noise is a significant source of noise pollution, disrupting urban environments with fluctuating sound. The existing research on traffic noise prediction predominantly focuses on statistical methods to identify significant predictors affecting noise levels. While these approaches offer valuable insights, they often lack the interpretability and adaptability needed for complex urban environments. The proposed framework is aimed at presenting the insights of explainable AI (XAI) for the regression analysis of traffic noise levels which is predicted with the help of advanced machine learning (ML) models such as K-Nearest Neighbor (KNN), Extreme Gradient Boosting (XGBoost), Long-Short Term Memory (LSTM) and Random Forest (RF). Statistical analysis of these models was tested with a performance matrix by utilizing a comprehensive traffic dataset of Dhanbad city that includes vehicle speed and categories of vehicle type. Notably, the RF model excelled over other models with an RMSE of 1.27 and [Formula: see text] of 0.94. The XAI model was developed with the base of RF regressor which records the highest [Formula: see text] score. The analysis revealed that the number of 2-wheeler vehicle categories is a key predictor of traffic noise levels. The finding of this study can act as an automated information system for the benefit of the urban planners and decision-making bodies to mitigate noise pollution effectively in mid-sized cities. It is worth mentioning that the primary purpose of employing multiple ML models (RF, XGBoost, KNN, LSTM) in this study is to conduct a comparative analysis and identify the most suitable algorithm for urban traffic noise prediction in a Tier-2 Indian city.