A Machine Learning Approach for the Assessment of Road Traffic Noise: Comparison of Regressors

In recent years, there has been an increasing use of Machine Learning and Artificial Intelligence methods in acoustical applications particularly in the development of predictive models for vehicular traffic noise. Such approaches differ from the original noise emission model in their principle of functioning, and they are gaining increasing attention for the possibility of describing uncommon road traffic scenarios. As a drawback, they usually require many data to be properly calibrated. In this contribution, four regressors (Multilinear Regression, Decision Tree, Random Forest, and Support Vector Regressor) and their performance in predicting vehicular traffic noise levels from data available in the literature are confronted. The results show that the implemented regressors can perform a good prediction of road traffic noise data, with a maximum mean error of -0.29 dBA and an MAE of 1.39 dBA. The differences between the regressors are due to the different statistical rules implemented, and yet they are not so prominent in identifying one regressor as the best over the others.