Traffic flow is an interesting many-particle phenomenon, with nonlinear interactions between the vehicles. This paper studies highway traffic and presents a general framework for traffic modelling as a stochastic hybrid system (with contin uous and discrete dynamics and some interactions between its components). The freeway is considered as a network of components, each component modelling a diffeerent section of the traffic network. A model is developed in order to design stochastic predictors of the flows. The different traffic modes and transitions between them are detected and analyzed in real data, collected by induction loop detectors from Belgian and Dutch freeways.