Traffic Flow Forecasting Neural Networks Based on Exponential Smoothing Method

This paper discusses a neural network development approach based on an exponential smoothing method which aims at enhancing previously used neural networks for traffic flow forecasting. The approach uses the exponential smoothing method to pre-process traffic flow data before implementing on neural networks for training purpose. The pre-processed traffic flow data, which is lesser non-smooth, discontinuous and lumpy than the original traffic flow data, is more suitable to use for neural network training. This neural network development approach was evaluated by forecasting real-time traffic conditions on a section of the freeway in Western Australia. Regarding training errors which indicate capability in fitting traffic flow data, the neural network models developed by the proposed approach was capable to achieve more than 20% of the rate of improvement relative to the neural network developed based on the original traffic flow data. Regarding testing errors which indicate generalization capability for traffic flow forecasting, the neural network models developed by the proposed approach was capable in achieving more than 8% of the rate of improvement relative to the neural networks developed based on the original traffic flow data.