APPLICATION OF TIME-SERIES ANALYSIS TECHNIQUES TO FREEWAY INCIDENT DETECTION

An approach for the automatic detection of freeway capacity-reducing incidents based on the time-series analysis techniques formulated by Box and Jenkins is suggested. An autoregressive integrated moving average model of the form ARIMA(0,1,3) has been recently developed to describe the dynamic and stochastic character of freeway traffic variables. This model is used to provide short-term forecasts of traffic occupancies and the associated 95 percent confidence limits. An incident is detected if the observed occupancy value lies outside the confidence limits of the corresponding point forecast. A total of 1692 min of occupancy observations associated with 50 traffic incidents that took place on the Lodge Freeway in Detroit are used in evaluating the algorithm performance. The algorithm detected all 50 incidents. The resulting false-alarm rate is 2.6 percent when constant parameters of the ARIMA model are used and it decreases to 1.4 percent with variable-parameter estimates. Furthermore, the average time lag to detection when constant- and variable-parameter estimates are used is 0.58 min and 0.39 min, respectively. (Authors)