Travel Demand Forecasting Under Interval Uncertainty

The travel demand forecast is the base of transportation planning, which directly affects the decision of policy-maker. However, in the traditional travel demand forecasts, uncertainty is usually ignored and deterministic point prediction methods are used which bring forth massive decision risks. In order to raise the reliability of forecasting result, the uncertainty should be taken into account in travel demand forecast.In this thesis, the problems about travel demand forecast models and algorithms under interval uncertainty are studied. Firstly, the uncertainty of travel demand is qualitatively analyzed to point out the main influence factors clearly. By contrast with four different kinds of uncertainty measurement, the advantage of interval number measuring the travel demand uncertainty is demonstrated. Secondly, a traffic assignment model under interval uncertain demand is established. Using interval analysis method, a robust shortest path algorithm is designed based on minmax regret principle and an interval successive average method is put forward to solve the traffic assignment model. Thirdly, considering uncertainty of interval link flow, an interval bi-level model is established to estimate interval OD matrix. Combining interval analysis with genetic algorithm, an interval genetic algorithm is designed by constructing new genetic operators which can dispose interval operations. Finally, the MATLAB program is developed to solve the models, and a simple example is used to test the above models and algorithms, including robust shortest path algorithm under interval impedance, method of interval traffic assignment and interval OD matrix estimation.The traditional forecast model of travel demand is improved by using interval analysis and considering demand uncertainty. The forecast result is a interval value which implies a variety of uncertainty and avoids decision-making risk in traditional point prediction ,thus enhancing the reliability of prediction. The policy-maker can make a more reasonable plan based on the results forecasted.