Traffic parameter estimation and highway classification: Rough patterns using a neural networks approach

Neural networks provide more accurate estimations of traffic parameters than conventional methods. This paper explores the possibility of using more sophisticated neural networks based on rough patterns for increasing the accuracy of estimations. A rough pattern is represented by upper and lower bounds of the input values. The paper compares four different data collection schedules and two different types of neural network architectures for estimations of average and peak traffic volumes as well as classification of highways.