Traffic flow prediction based on local mean decomposition and big data analysis

Abstract

In the era of the big data, the accurate prediction of real-time traffic flow is essential to making rational decisions on travel time, cost and route. To forecast traffic flow accurately, this paper firstly analyzes the features of traffic data, and proves that the traffic data collected from an overpass are self-similar. For simplicity, the long-term correlation (LTC) time series of the traffic data were decomposed into short-term correlation (STC) product functions (PFs) through local mean decomposition (LMD). On this basis, a traffic flow prediction model was developed based on the generalized autoregressive conditional heteroskedasticity (GARCH) model. Simulation results show that our model was more accurate in predicting traffic flow than the original GARCH and the autoregressive integrated moving average (ARIMA) model. Therefore, this research provides a suitable tool for the prediction of traffic flow.