Database Reference
In-Depth Information
phase of the analysis is crucial, as only after assessing the correspondence of
the preliminary analysis results with the ground truth can we proceed with
the mobility knowledge discovery steps having the guarantee that the results
will represent real mobility patterns. In Chapter 10 , dedicated to car traffic
monitoring, we will see an example of this validation process on the Milano
data set.
Trajectory Reconstruction and Preparation
As explained in the previous chapter, the data mining algorithms apply to the
concept of trajectory: but which trajectory definition? It is simply the ordered
sequence of observations of the user's history? Or a subsequence represent-
ing the movements between stops? And how to define and compute a stop?
Answering these questions is crucial and affects deeply the results of the knowl-
edge discovery process. For example, if we are interested in frequent paths
followed by a certain number of users we need to consider T-pattern applied
to the whole user history as single trajectory, so that the support of a single
pattern will be the number of users that follow that path. Alternatively, if we
are interested in the usage of certain frequent paths then we do not need to dis-
tinguish between distinct users. As a consequence, the concept of trajectory to
be mined becomes the subsequences of user trajectories delimited by two stops
as described in Chapter 2 . There are several ways of reconstructing trajectories
considering different constraints and thresholds thus leading to different sets of
trajectories. In M-Atlas we can perform this operation with the data constructor
statement .
CREATE DATA <trajectory_table> BUILDING MOVING_POINTS
FROM (SELECT userid, longitude, latitude, datetime
FROM <raw_observation_table>
ORDER BY userid, datetime)
SET MOVING_POINT.<constraint_name> = <value> AND ...
The syntax of queries in M-Atlas extends the standard SQL. In this query we see
a CREATE DATA operation building a new kind of data to be stored in the data-
base from a pure relational table. As we can see, a new trajectory table is built
from the raw observations using trajectory reconstruction parameters expressed
in the constraints . Some examples are MAX_TIME_GAP or MAX_SPEED which
realize the two constraints described in Chapter 2 . These values depend on
the application and their values have to be carefully chosen because they
affect all the subsequent analysis. Examples are MAX_TIME_GAP = 30 min
or MAX_SPEED = 5 km/h to cut trajectories where there is a temporal gap of
30 minutes or a max speed of 5 km per hour, respectively.
Search WWH ::




Custom Search