| description abstract | Different parameter results are produced using various methods with the same sample, which follows the lognormal distribution. Therefore, massive historical data of an expressway toll are collected to fit the lognormal distribution of the travel times of cars and trucks. Maximum likelihood estimation is compared with least squares estimation based on five indexes: the sum of squared errors (SSE), the SSE of the cumulative distribution function, the coefficient of determination, the relative error of the skewness coefficient, and the relative error of the kurtosis coefficient. Experimental results show that the maximum likelihood estimator is more efficient than the least squares estimator in fitting the cumulative probability distribution. The unit distance travel time reliability model for the section is established, and the parameter estimation method and the appropriate threshold are provided. Finally, the proposed model is validated using the Shaanxi expressway network in a case study. The results show that travel time reliability is affected by travel distance and departure time. In addition, departure time more significantly affects the travel delay of cars than that of trucks. | |
