The purpose of this dissertation is to investigate the potential in using a back-propagation neural network for a traffic demand(volume) forecasting in the isolated intersection. Traffic demand forecasting is very important to a traffic planning and a traffic control in the isolated intersection.
The necessity to adopt compromise signal settings with fixed time signals handling fluctuating volumes immediately reveals their major disadvantage. They are unable to adjust themselves to changing traffic conditions. Vehicle-actuated signals, within certain limits, do not suffer from this drawback. Vehicles approaching an intersection register there presence by actuating a demand signal through a detector, which is a sensing device, linked to a controller. Basically, the controller is an electronic timer which governs the cycle time and changes the signal aspects in response to traffic demands. But they are passive control systems to given traffic conditions and inefficient to unexpected traffic demands in the isolated intersection. Therefore, traffic demand forecasting is necessary for the efficient traffic control in the isolated intersection.
Traffic demand(volume) in the isolated intersection can be forecasted by the interactions between traffic volumes of isolated intersections in a road network. There are many complex patterns of traffic volumes between isolated intersections and it is very difficult to analyze them mathematically. But using the outstanding pattern recognition ability of the neural network, it is able to learn these traffic patterns and perform the desired mapping on traffic patterns it has never encountered during learning (training).