The objective of this paper is to review techniques that are used to predict travel flow and apply some of the techniques on field data from Melbourne's freeways to evaluate the accuracy and robustness of the techniques. The techniques used in the evaluation are: (1) Regression, (2) Historical average, (3) ARIMA, and (4) SARIMA. The basic premise behind the historical data based algorithm is that traffic patterns are seasonal. In other words, a knowledge of typical traffic conditions on Tuesday at 5:30pm will allow one to predict the conditions on any particular Tuesday at 5:30pm. Although these algorithms perform reasonably well during normal operating conditions, they do not respond well to external system changes such as weather, special events, or modified traffic control strategies. ARIMA (Auto Regressive Integrated Moving Average) is a statistical based method of time series analysis. It is based on the premise that the knowledge of past values in a time series is the best predictor of the variable in question. In other words, the ARIMA model can produce accurate short term forecasts based on a synthesis of historical patterns in data and does not assume any pattern in the historical data of the time series. If the time series is seasonal (a series that has a pattern that repeats itself over fixed time intervals), a seasonal ARIMA (SARIMA) can be applied to handle these specific aspects of the time series. Results from the comparative study are presented in this paper. (Author/publisher) For the covering entry of this conference, please see ITRD abstract no. E205861.
Samenvatting