The origin-destination travel demand matrix is one of the most basic and important information sources in the planning, management and control of urban transportation networks. The traditional methods of estimating the demand matrix are usually used for long-term periods (every ten years) due to their time-consuming, manpower and high cost. Therefore, it will be important to use alternative methods and sources of information in order to achieve this matrix. This research presented a method for estimating the time-dependent origin-destination demand matrix in the Isfahan city traffic network using automatic license plate recognition systems data and data mining approach. According to the approach of extracting all the required information from the studied data to estimate the time-dependent travel time, an designed algorithm estimates the travel time between each pair of systems every hour of the day by spatial and temporal clustering using genetic algorithm. Then using the proposed algorithm, the path flow between each pair of cameras is estimated by dividing the complete reconstructed vehicles trajectory into several trips based on the time and place of passage. Finally, the origin-destination travel demand matrix is estimates by assigning the path flows between each pair of cameras to the modified traffic zones. This method was implemented on the information of full day of the network and also the estimated matrix was compared with the developed matrix of comprehensive transportation studies of Isfahan city. The results show that the mean absolute percentage error (MAPE) of the estimate is 38.4% for trip generation and 50.6% for trip attraction of zones. The low coverage rate and number plate reading rate of license plate recognition systems are among the effective factors in increasing the travel demand matrix error. Also, considering the spread of Corona disease in the investigated time period, the eva‎luations show its impact on the change of travel goals, areas demand and as a result, the cause of the difference between the estimated travel demand matrix and the transportation studies matrix. In the presented method, in addition to needing no previous matrix and no affecting by its patterns, it is possible to obtain new patterns in the shortest time when the previous matrix has undergone changes.

حسين حق شناس

احمدرضا طالبيان، احسان سيد ابريشمي