Time Series Models for Freeway Incident Detection

An autoregressive integrated moving average model of the form, ARIMA (0,1,3), is presented for describing the stochastic and dynamic behavior of freeway traffic volume and occupancy observations. The model is applied to the detection of freeway capacity-reducing incidents through the sudden and pulsed changes they generate in traffic stream time series data. An incident is detected if the observed value of traffic occupancy lies outside the probability limits constructed two standard errors away from the corresponding point forecast. The developed ARIMA occupancy algorithm has been found superior to the exponential and the California algorithms in terms of detection rate, false alarm rate, and mean time lag to detection. The analysis is based on surveillance data recorded at the Los Angeles, Minneapolis, and Detroit freeway systems during the afternoon peak periods.