Research on short-term traffic flow prediction model based on threshold autoregression

Traffic congestion is a problem faced by cities, and it is urgent for solving this issue. The real-time and accuracy of traffic flow prediction directly affect the efficiency of traffic flow guidance system，which is a hot issue of intelligent transportation system research. The time series of the same traffic flow has different characteristics in different time buckets, so the traffic flow sequence is divided into subsequences with different characteristics by the accumulative departure method. A prediction model of short-time traffic flow sequence based on threshold autoregression (TAR) is established by taking the above time as the threshold value. The real-time traffic flow time series of city expressway in Beijing is taken as the object of empirical study. The results show that the average error and mean absolute error of the model are -0.38% and 7.38% that are lower than 3.08% and 7.73% of ARIMA prediction model when the prediction step size is 1. When the prediction step size is 5, the average error and mean absolute error are -0.71% and 14.78% that are lower than 5.40% and 17.44% of ARIMA prediction model. The simulation results demonstrate that the algorithm has better prediction accuracy, which can verify the feasibility and effectiveness of the algorithm in the prediction realm.