Recently, several urban areas are trying to mitigate the environmental impacts of traffic, where noise pollution is one of the main consequences. Thus, studying the determinants of traffic-related noise generation and developing a model that predicts the level of noise by controlling the influencing factors are crucial for transportation planning purposes. This research aims at utilizing the response surface method (RSM) to develop a robust statistical prediction model of traffic-related noise levels and optimize different traffic characteristics’ ranges to reduce the expected noise levels. The results indicate that the rate of L_eq increase is higher at traffic flow values less than the 1204 veh/h. The interaction effect of flow-speed and flow-heavy vehicle percentage pairs shows that L_eq has peak values around 45.8 km/h and 28.71%, respectively, with almost symmetric value distribution about those center points. The main effects study indicates a direct effect of traffic flow, speed, density, and traffic composition on roadside noise levels. The prediction model has good representativeness of observed noise levels by predicted noise levels as the model has a high coefficient of determination (R² = 95.87% and R² adj = 92.26%) with a significance level of 0.0036. Then, the research presents a methodology to perform an optimization of the roadside noise level by defining traffic characteristics that can keep the noise level below 65 dB(A) or minimize noise level. Decision-makers could use the proposed method to control the roadside noise level.

Promoting public transport can increase the role of transport in sustainable development. Thus, studying the determinants of choosing public transport by travelers is crucial for transportation planning purposes, where developing an accurate model can help in examining any proposed scenarios. This paper aims to develop a multivariable regression model to describe the willingness to use public transport (W) represented as the percentage of people who use public transport in United States cities. First, census data of socio-economic and demographic characteristics are analyzed to identify significant factors for W to develop the model. Then, the regression technique is utilized to develop the model. The model is statistically assessed, in which the significance of all independent variables is examined and represented by a p-value. Moreover, the correlation between variables is examined. Then, the most statistically appropriate model for W is identified based on a set of performance measures such as coefficient of determination, average error, geometrical mean, Theil's inequality coefficient, and frictional bias. Finally, sensitivity analysis is conducted to assess the elasticity of the model to changes in the significant variables by considering a 10% change (increase or decrease) in the average of each variable.