An inference method for a level of traffic congestion using neural network.

In this paper, an inference method for determining a level of traffic congestion in terms of the confidence level of sensors is developed using the neural network. Specifically the level of traffic congestion at a target link is accurately estimated by observing traffic flows of the neighboring links. Inputs of the neural network consist of the level of neighboring traffic congestion and the confidence level of sensors. The output of the neural network is the level of traffic congestion of the target link. This paper addresses the key issues on inferring the level of traffic congestion of the link where any sensor is not installed. The performance of the proposed inference method is evaluated using traffic simulation data obtained by NETSIM. (A*)