Comparison of regression model and artificial neural network model in noise prediction in a mixed area of Dhaka City

Abstract

The equivalent noise levels regularly exceed acceptable limits within Dhaka city, the capital of Bangladesh, especially in the mixed urban areas (where trips are generated to serve commercial, residential, and industrial demands). The study aims to assess the noise level in mixed urban areas, build noise prediction models and allow scopes for ensuring sustainable environmental management. Two traffic noise prediction models were assessed: a regression model and an artificial neural network (ANN) model to predict the equivalent noise level (Leq). Traffic and noise level data were collected from two mixed urban areas, statistical analyses were performed to describe the existing trends and to evaluate both model’s responses in predicting equivalent noise level (Leq). The ANN model (coefficient of determination: 0.82) showed better performance than the regression model (coefficient of determination: 0.70). The predicted equivalent noise levels from the ANN model were compared to acceptable limits to display the extent of noise pollution using GIS. The traffic noise models can assist in environmental impact assessment to protect the communities susceptible to the adversities of noise pollution.