ARIMA model for traffic flow prediction based on wavelet analysis

As the traffic flow has the features of nonlinear and strong interference, it has different characteristics in different time-frequency spaces. Firstly, this article uses the wavelet analysis method, decomposes a group of original traffic flow signals containing summarized information into series of time sequence signals that have different characters, then makes use of good linear fitting ability of the ARIMA model processes the wavelet analysis time signal through the ARIMA model. Using matlab and SPSS, the measured traffic flow data were analyzed verified. Experiment results show that the way of combining the wavelet analysis with ARIMA model can reduce the prediction error effectively, and improve the forecasting accuracy by about 80%, this way has high feasibility.