Potential assessment of an ellipsoidal neural fuzzy time series model for freeway traffic prediction

Abstract

Forecasting of traffic flow is one of the most important approaches to control the capacity of highway network efficiently during peak flow periods. Therefore, many emerging methods have been designed to predict traffic flow of freeways. However, the ellipsoidal neural fuzzy model, originally developed for control and pattern recognition problems, was seldom used in forecasting traffic flow. The aim of this study is to investigate the potential of ellipsoidal neural fuzzy model in predicting highway traffic. Monthly traffic data at Tai-Shan tollgate of a freeway in Taiwan are collected to depict the performance of forecasting models. Three other neural network models, namely back-propagation neural networks (BPNN), and radial basis function neural networks (RBFNN) and general regression neural networks (GRNN) models are used to predict the same traffic data sets. Simulation results reveal that the ellipsoidal neural fuzzy time-series (ENFTS) model is superior to the other models. Therefore, the ENFTS is a feasible and promising approach in predicting freeway traffic. © Springer-Verlag Berlin Heidelberg 2007.