Short-term prediction of a nonrecurrent road traffic using multifractal tools

We present a first approach for short-term forecasting of traffic flow data using multifractal analysis. The acquisition time is assumed very small (1 minute in our case), which yields highly irregular data well suited to multifractal analysis. We perform predictions with different horizons (from 1 minute to 15 minutes) and come up with satisfactory results. We then present a possible improvement to this approach based on multinomial measures and results on multifractal correlations