ANALYSIS OF FREEWAY TRAFFIC TIME-SERIES DATA BY USING BOX-JENKINS TECHNIQUES

Abstract

This paper investigated the application of analysis techniques develoepd by Box and Jenkins to freeway traffic volume and occupancy time series. A total of 166 data sets from three surveillance systems in Los Angeles, Minneapolis, and Detroit were used in the development of a predictor model to provide short-term forecasts of traffic data. All of the data sets were best represented by an autoregressive integrated moving-average (ARIMA) (0,1,3) model. The moving-average parameters of the model, however, vary from location to location and over time. The ARIMA models were found to be more accurate in representing freeway time-series data, in terms of mean absolute error and mean square error, than moving-average, double-exponential smoothing, and Trigg and Leach adaptive models. Suggestions and implications for the operational use of the ARIMA model in making forecasts one time interval in advance are made. /Author/