The analysis and design of traortation systems require the estimation of present demand and the forecasting of future demand. These estimates and forecasts can be obtained using a variety of information sources and statistical procedures. Information about Trip demand i The study area Can be save in origin destination matrix. The components of this matrix shows the number of trips between the area of the region. Information of this matrix can be earned in two ways, direct and indirect. Direct methods For data collection includes questionnaire, interview and observation methods, from the entire population. In the last few decades, Due to time and cost too much to gather information for people traveling in an area in directed method, indirect methods of estimating travel demand matrix has been of interest to researchers. Some of these methods Using traortation network links traffic counts adjast and update an initial origin destination matrix. On of the directed method is statistical infrence that includes maximum liklihood, generalised least square, and bayesian infrence approachs. In this study, one of the statistical approach, called Bayesian inference, for using in large traortation network is researched. Methods that is used for large traortation network have a common feature, and all of this method using a bilevel structure. In upper level of these problems, origin Destination matrix is estimated, and in a lower level, origin destination matrix is assigned to Traortation network. In this thesis for upper level a computational structure for Estimation of origin destination matrix using bayesian network is applied that can estimated origin destination matrix for large traortation network in a short time, and in a lower level problem trip demand matrix estimated in each iteration assigned to traortation network. In this research For traortation operations in large traortation network we used traortation software EMME/3. Then for evaluation of proposed model, method is tested for Sioux falls network And its results with using statistical software is compared with spiess gradiant method. Finally method is tested on Isfahan Traortation network. Keywords: Origin-destination matrix, traffic assignment, Bayesian Networks, Bi-level algorithm
