Maximum entropy models for mode split forecasting

The distributions of some random variables of value of travel time data are quite complicated and difficult to determine by using ordinary statistical models. Two maximum entropy functions are borrowed to describe value-of-time (VOT) distributions from different sample size. The traditional VOT-based mode split model is improved by the use of maximum entropy principle. A new algorithm is proposed for estimating parameters of the maximum entropy quantile function. The application of the models and the algorithm is illustrated via 2006 travel survey data from the city of Guangzhou, China. Empirical evidence demonstrates the efficiency of new models and algorithm.