GCBN: A hybrid spatio-temporal causal model for traffic analysis and prediction

Abstract

In this paper we propose a binary Bayesian network to model the speed variations for traffic speed prediction. Comparing to continuous graphical models, firstly, our method reduces the complexity of the model. Secondly, we use Granger causality test to determine the structure and parameters of the Bayesian network. Experiments on large GPS data of vans in the freeway network illustrate the good performance of our model. © 2013 Springer-Verlag Berlin Heidelberg.