Multi-Scale Analyses and Combining Forecasts for Traffic Volumes

In view of that the current models are almost combining models of different kinds of forecast methods and the absence of substantial research on multi-scale combining in the context of traffic forecasts, we proposed a method that in multiple scales combines multiple forecasts based on prediction approaches of the same category. The method combines multi-scale analyses with exponential smoothing forecasting and its weights are assigned according to the performance of the various scales' forecasts. Using four years' actual data, we in two scales demonstrated it by forecasting annual traffic volumes for an expressway in China and the application verified its accuracy.