Database Reference
In-Depth Information
(a) (b)
Figure 10.13 Distribution of presence: (a) with predicted trajectories, (b) with the real
trajectories. As highlighted on (a), the predictor is able to correctly guess the most dense
locations (green circles), though it introduces some false positives (red circles). (See color
plate.)
data set that includes trajectories from five working days (from Monday, July 5
to Friday, July 9) restricted to the morning peak hours (8 A.M.-10 A.M.). This
selection resulted in about 10,000 trajectories for the training set. Then, we
selected, as test set, the trajectories of Monday, July 12th (in the same temporal
interval), leading to a total of around 4,000 trajectories. From them, the algo-
rithm was able to predict the next location of about 3,000 trajectories focused
on 29 regions. Five of them contain more than 150 trajectories. Scaled to the
global number of circulating vehicles this corresponds to about 7,500 vehicles
predicted to converge to these areas in the two-hour interval. Figure 10.13 reports
the results of the prediction compared with the ground truth obtained by com-
puting the density map of the real GPS trajectories moving during the predicted
period.
It is worth pointing out that the interpretation of the predicted zones suggests
further, deeper analysis. Indeed, the dense regions do not necessarily indicate
traffic problems in those areas. These regions represent dense movements of cars,
which can hint the possibility of traffic jams or congestions. Further analysis,
focused on these specific areas, are needed to have a more precise indication of
possible traffic problems.
10.4.6 Borders of Human Mobility
Here, we address the problem of finding the borders of human mobility at the
lower spatial resolution of municipalities or counties. The aim of discovering
borders at a mesoscale is motivated by providing decision-support tools for
policy makers, capable of suggesting optimal administrative borders for the
Search WWH ::




Custom Search