Performance Evaluation of the ANN and ANFIS Models in Urban Traffic Noise Prediction

Abstract

Urban traffic is currently considered as one of the noise sources in the city of Erzurum province. As a disturbing sound, noise has several effects on environmental health and should be determined precisely in order to prevent or mitigate its effects in daily life in cities. This study aims to predict traffic noise in urban areas using with two well-known methods, artificial neural networks (ANN) and adaptive neuro fuzzy inference system (ANFIS). In order to compare the results of the methods same input data set consisting of total number of hourly vehicles, heavy vehicles, their average speeds were used in prediction process and 10 percentile exceeded sound level (L-10) were produced as the output of the models. Results of the study outlined that ANFIS model operated better than ANN model for the prediction of the noise originated by the urban traffic based on the statistical results, R-2 of ANFIS and ANN models were determined as 0.91 and 0.81 respectively. Additionally, this study concluded that the prediction of traffic noise under heterogeneous traffic which is mostly complicated with based on vehicle number, driver behaviors that causes to irregular pattern of factors.