Vehicular-cloud simulation framework for predicting traffic flow data
Sahraoui Abdelatif, Makhlouf Derdour, Ahmed Ahmim, Philippe Roose
Abstract
The traffic flow prediction has become an important process tailored with the exponential development of cities and the transportation systems. The main purpose of the prediction task is to improve the logistic services and reduce the cost of the road congestion. In this paper, we propose a vehicular-cloud simulation framework with a layer of traffic cloud services to predict accurate traffic flow data. Learning of supervised traffic flow data from several data sources is the core of these services. Particularly, we focus on a particular type of dependency (i.e., monotone dependency) between the learning traffic inputs and its responses. The learning algorithm we propose aims to solve the regression problem by predicting values of a continuous measure. The accuracy of the proposed cloud services have been tested under congestion conditions, where the results show better performances over short periods and daily forecasts.