The analysis of network traffic flows needs origin-destination matrix information. This matrix consists of the origin-destination data of the traortation network which can be obtained via two direct and indirect methods. Since construction of the origin-destination matrix using direct methods is so much costly and time consuming, hence using indirect methods has received more attention. One of the most important indirect methods is estimation of origin-destination matrices by counting traffic of some of the links. The accuracy of the origin-destination matrix is related to some data such as the initial origin-destination matrix and the links chosen for counting. On the other hand the budget for counting the links is limited, thus selecting links which give the most information is so important. The estimation and upgrading of the origin-destination matrix has been investigated extensively but the location of the traffic counting points has received very limited attention. One of the accurate methods for selecting optimal traffic counting location that has received more attention in recent decades is using the Bayesian networks. The problem of this method is its long using time for large networks. In this thesis two solutions are presented for using this method for large networks. These solutions are omitting the unimportant links and decreasing the number of origin-destination pairs. Also the new indicator for selecting the links is presented. At last this method is installed for Isfahan traortation network. It is obvious that by much changing of some zones’ attractiveness for traveling, this method has some constraints in installation
