Mathematical, statistical and neural models capable of predicting LA,max for the Tehran-Karaj express train

This paper presents mathematical logarithmic, statistical linear regression, and neural models capable of predicting maximum A-weighed noise level (L_A,max) for the Tehran-Karaj express train. The models have been developed upon the basis of the measurements from sampling locations at distances of 25 m, 45 m, and 65 m from the centreline of the track and at a height of 1.5 m. In the next step, the predictive capability of the models have been tested on the data associated with the sampling locations, situated, respectively at distances of 35 and 55 m from the centreline of the track at a height of 1.5 m. The non-parametric tests i.e. two-related samples Wilcoxon, and two-independent samples Kolmogorov-Smirnov, carried out, respectively for training and testing steps, indicate satisfactory results. In the final step the non-parametric k-related samples Friedman test detects no significant differences amongst the absolute testing set error of the models.