Urban Road Traffic Noise Modelling In The Mid-Sized City Of Agartala: Using Multiple Linear Regression (MLR) And Artificial Neural Network (ANN) Techniques

Abstract

Road traffic noise is a significant source of environmental pollution in urban areas, with increasing traffic leading to elevated noise levels that adversely affect city life. As a result, ongoing research globally focused on controlling and mitigating traffic noise. This problem is particularly severe near intersections, where inadequate planning and the absence of noise reduction strategies exacerbate the situation. Therefore, this study develops a traffic noise model for the mid-sized city of Agartala to evaluate vehicular noise levels at intersections. It evaluates two noise prediction models, multiple linear regression (MLR) and artificial neural network (ANN) to estimate the equivalent sound level (Leq), using variables like total hourly vehicles, percentage of heavy vehicles, vehicle speed, road width, temperature, and humidity. The ANN model outperformed the regression model, achieving an r of 0.956, R² of 0.9139, MSE of 1.74, RMSE of 1.32, MAPE% of 1.41, and an accuracy (± 1 dBA) of 83.87%. In contrast, the regression model had an r of 0.905, R² of 0.8187, MSE of 3.57, RMSE of 1.89, MAPE% of 2.20, and an accuracy (± 1 dBA) of 48.39%. ANN is a highly effective tool for modelling traffic noise. The study’s outcomes could serve as valuable resources for noise modelling consultants and urban planners mapping traffic noise in mid-sized urban areas.