Understanding Human Mobility Using Mobility Data Mining - Mobility Data - page 132

Data Manipulation
Before the execution of a data mining algorithm the analyst can manipulate
(e.g., select the data in a particular area or period) or transform the data (e.g.,
anonymize for privacy reasons). To these purposes, the system provides a rich
set of operations: the relational statement represents the creation of a relation
between two objects applying a predicate while the transformation statement
modifies the original data according to a transformation function (or algorithm).
To better understand the relation statements and transformation statements we
present two examples. The first one is the relation between a trajectories table
and a temporal period table, which computes the temporal distribution of the
movements:

CREATE RELATION <relation_table> USING INTERSECT
FROM (SELECT t.id, t.object, p.id, p.object
FROM <trajectories_table> t,
<time_periods_table> p)

The objective of this query is to create a new table where the trajectories are
intersected with a temporal period. This is useful in the analysis process when
the data to be mined have to be selected based on space and/or time, as in the
example above. All the spatio-temporal operators embedded into the system -
such as INTERSECT - assume a different meaning according to the types of data
to which they are applied. For example, the INTERSECT operator when applied
to two trajectories becomes a spatio-temporal intersection and this operation
returns true when the two moving users are in the same place at the same time.
The second example is a transformation operation that builds a new set of
trajectories to be mined. A classic example is the anonymization of trajecto-
ries, where the initial data set is transformed to guarantee a certain degree of
anonymization of the trajectories. The main idea is that, in the anonymized
data set, each individual is indistinguishable from other k 1 individuals, as

detailed in Chapter 9 . However, for explaining the TRANSFORMATION operation
it is important to point out that the original data set is changed into a new one with
some properties that, in this case, guarantee the anonymity of the individuals.

−

CREATE TRANSFORMATION <trans_table> USING K-ANONYMITY
FROM (SELECT * FROM <trajectories_table> t)
SET K-ANONIMITY.K = <k_value>

7.2.2 Data Mining

Naturally, themining step applies themining algorithm. However, several actions
can be taken during the mining step in order to make the knowledge discovery
more effective. Moreover, models can be further manipulated and combined.