Hybrid Traffic Flow Forecasting Model Based on MRA

The presence of complex scaling behavior in traffic makes accurate forecasting of traffic a challenging task. This paper proposes a multi-scale decomposition & reconstruction approach for real-time traffic prediction. The proposed scheme combines the superior characteristics of wavelet neural networks, ARIMA and MRA. This multi-scale decomposition and reconstruction approach can better capture the correlations within traffic flows caused by different mechanisms, which may not be obvious when examining the raw data directly. The proposed hybrid prediction algorithm is applied to real-time traffic data from a large metropolitan area. It is shown that the proposed algorithm generally outperforms traffic prediction using a single prediction model approach and gives more accurate results.