Current vehicle navigation systems receive congestion information via RDS-TMC, to the
extent that congestion is detected. A better quality of guidance would be possible, however,
if short-term prediction of future congestion were available beside information on current
congestion.
Congestion is thought to be spatially and temporally correlated on the road network. A
notable phenomenon is that the congestion on a link is affected by its immediate past traffic
condition and the number of its neighbouring links. A time component defines how traffic
conditions are related in the temporal dimension, and a spatial component defines how the
traffic conditions on neighbouring links have affected each other in the spatial dimension.
This paper explores the possibility of incorporating spatial and temporal effects in congestion
prediction using a zero-inflated Poisson model. A year‟s worth of TMC data from a
particular area of London have been collected and input into the model for this purpose, and
the results are presented in this paper.