Trajectory Data Warehouses - Mobility Data - page 81

WHERE projPres = favg(atperiod(atregion(p,d.geometry),21/1/2012))
AND (area(defspace(atregion(at(l,'Residential'),
d.geometry)))/area(d.geometry)) >= 0.7
In this query, the time-dependent field Presence is projected to the geometry
of the district and to the date January 21, 2012, and the favg operator is applied
to compute the average of presence across hours of the day. The resulting
nontemporal field is kept in the variable projPres . On the other hand, the
nontemporal field LandUse is projected to residential zones and to the geometry
of the district, and the correspoding region is divided by the area of the district
to verify the 70% condition specified in the query.
4.7 Conclusions
We have discussed data warehousing techniques that, in the presence of trajec-
tory data, help to improve the decision-making process. For this, we defined
the notion of trajectory data warehouses (TDW) as a particular case of spatio-
temporal data warehouses, where trajectories can be represented both as mea-
sures and dimensions. By means of a running example we showed how a
TDW can be modeled, designed, and queried, in order to deliver an aggregated
view of trajectory data. In addition, as a particular case study, we discussed
the GeoPKDD TDW, where facts contain aggregated trajectory measures
instead of the trajectories themselves. Finally, we showed that representing the
GeoPKDD TDW as a collection of continuous fields, one for each measure,
provides additional possibilities for analysis.
4.8 Bibliographic Notes
Basic data warehousing concepts can be found in the classic topic by Kim-
ball ( 1996 ). This chapter is based on previous research work on spatio-temporal
data warehousing and continuous fields performed by the authors ( Vaisman and
Zimanyi , 2009a , b ). Hierarchies in OLAP are studied, among other works, in
Cabibbo and Torlone ( 1997 ). MultiDim, the conceptual model we use in this
chapter, was introduced in Malinowski and Zimanyi ( 2008 ). The query language
we use throughout the chapter is based in the classic relational calculus with
aggregate functions introduced by Klug ( 1982 ). The data type system follows
the approach of Guting and Schneider ( 2005 ). The view of continuous fields
as cubes was introduced in Gomez et al. ( 2012 ). The GeoPKDD TDW, its
associated ETL process, and the double-counting problem during aggregation
are studied in Orlando et al. ( 2007 ). A good discussion on TDW is presented