Database Reference
In-Depth Information
FROM (SELECT id, trajectory FROM TrajectoryTable)
SET SPECIMENS_CLASSIFIER.SPECIMENS =
(SELECT * FROM WednesdaySpecimens) AND
SPECIMENS_CLASSIFIER.METHOD = ROUTE_SIMILARITY
This approach to the management of the mobility knowledge discovery pro-
cess allows the interoperability of models and data, and it also provides a clear
tool to summarize and formally define the analytical process.
10.4.3 Access to Key Mobility Attractors
To understand how users access big mobility attractors, we focus on the travels
ending in a specific parking lot of the city. An advanced knowledge of the
dynamics of use of a parking lot allows the mobility agency to plan specific
fares or to notify the users of extraordinary events or interruption of services.
For this case study we have selected the parking lots of the Linate Airport.
Figure
10.9
shows the set of trajectories that start in Milan and end in the
airport parking lot, selected by means of a OD matrix selection. It is evident
that vehicles start from a broad diversity of locations, but converge toward the
parking lot. Our goal is to characterize the typical behaviors of vehicles when
approaching the attractor, a task that cannot be directly addressed by clustering,
due to the fact that clustering generally works at the level of whole trajectories,
while the behaviors might emerge just on shorter subtrajectories. Also, simply
predefining a set of directions of approach and counting how many trips reach
the attractor from each of them answers our request only partially, as we want
to characterize behaviors, which might include not only incoming directions
but also particular paths followed (e.g., common shortcuts or detours). As an
example, we focus here on frequent segments of trips that are followed by a
significant volume of vehicles, a feature that can be directly detected by mining
trajectory patterns (see Chapter
6
). We recall that trajectory patterns describe
sequences of regions that appear frequently in the data, together with their typical
transition times. Figure
10.9
b is a visual summary of the trajectory patterns that
are supported by at least 5% of the travels to Linate. As we can see, they allow us
to characterize the three main routes to approach the attractor, together with the
different travel times. Figure
10.10
focuses on the three most frequent trajectory
patterns. Observe how the trajectory patterns approaching the airport from north
are longer than those from south, highlighting that the northern travels tend to
concentrate on the outer ring earlier than the southern travels, which instead
use a small segment of the ring. This behavior suggests the presence of more
alternative routes to get to the proximity of the airport from the south and city
center than from the north.
