Document Type : Research Article
Civil Engineering School Iran University of Science and Technology
School of Civil Engineering, Iran University of Science and Technology
Iran University of Science and Technology
School of Civil Engineering, Islamic Azad University Tehran Science and Research Branch
This study intends to determine the most appropriate distribution for modeling time variability. It also aims to explore the effects of the time of day and the length of the analysis time interval on the type of the best-fit probability distribution function. To this end, four analysis time intervals of different lengths ranging from five minutes to three hours are considered. Subsequently, for each analysis time interval, travel time data collected at different times of day are fitted to 12 common probability distribution functions. The Akaike Information Criterion is then used to evaluate the goodness of fitting and to rank the probability distribution functions. The results of this study indicate that the Gaussian mixture distributions are superior to single-component distributions to represent travel time distribution. In addition, as the length of the analysis time interval decreases, single-component probability distribution functions can better estimate the distribution of travel time observations. Among single-component probability distribution functions, the generalized extreme value distribution, the burr distribution and the log-normal distribution provide the best fit to the travel time data. The results of this research also show that the type of the best-fit probability distribution function does not change significantly over the time of day.