The detrended fluctuation analysis method is applied to analyse the complexity of traffic flow time series and the scale index is obtained, which is a parameter to depict the long term evolutionary behaviour of the mixed traffic flow time series. According to the changes of the scale index, it is shown that the traffic flow time series has long-range correlation characteristics or short-range correlation characteristics. It is found that the traffic flow series shows long-range correlation when the density ρ12; and when ρ1 or ρ>ρ2, the time series displays short-range correlation. That is to say, the vehicle density plays an important role in the changes of the scale index. Additionally, we also found that the ratio of slow vehicles have influence on the changes of the scale index.