Application of Wavelet Time Series Analysis in Short-Term Traffic Volume Prediction

Abstract

Short-term traffic volume forecasting is a key component of many intelligent transportation and traffic control systems. The Autoregressive Integrated Moving Average (ARIMA) models have been widely studied and used for short-term traffic volume prediction by many researchers. However, the fluctuations due to local noise make it difficult for ARIMA models to obtain high prediction accuracy. This paper investigates the use of Discrete Wavelet Transform (DWT) to improve the performance of ARIMA models for short-term traffic volume prediction. The paper first uses the DWT to denoise the original traffic volume data such that less fluctuating data are obtained. An ARIMA forecasting model is then fitted based on the denoised data series. Real-world data from Interstate 80 in California are used to test and compare this proposed wavelet ARIMA approach with ARIMA models.