GeoBI Architecture Based on Free Software: Experience and Review - Geographical Information Systems: Trends and Technologies - page 254

dimensions. Dimensions are associated with fact relationships(s) and may
be of one level, e.g., Cancer type , Gender , and Time , or may have several
levels forming hierarchies. For example, the District dimension includes
two hierarchies (with names placed in ovals): one called Administrative
division composed by District , County , and Province levels, and another
called Health division formed by District , Health area , and Region levels. To
indicate that some levels are spatial, a pictogram next to the level name is
placed, e.g., multi-polygons in Fig. 4, since many Costa Rican districts
are formed by islands. The relationships between hierarchy levels indicate
the many-to-one cardinalities, i.e., a district belong to one county and a
county may be related to many districts. Other cardinalities can also be
expressed, if needed. Notice that some levels include only one attribute (e.g.,
the Gender level). This representation on the conceptual level is important
since it allows users to better understand what kind of analysis s/he can
perform, e.g., considering population according to gender. As we will see
later, during the implementation, this level does not need to be represented
by a separate table.
By default, measures are considered additive (e.g., Births and Deaths
in the Demographics fact relationship in Fig. 4). To indicate that measure is
semi-additive (e.g., Population ) or non-additive (e.g., Incidence rate ), symbols
+! and + are used, respectively.
The MultiDim model allows sharing dimensions between different fact
relationships as can be seen in the Fig. 4 for the Time , Gender , and District
dimensions. Notice that the Cancer fact relationship in this fi gure includes
an additional Cancer type dimension; as a consequence, the level of details
(granularity) is smaller than in the Demographics fact relationship.
Furthermore, the conceptual level design requires the inclusion of an
abstract specifi cation of the required mapping between the source and SDW
data. Table 1 includes some examples of these mappings that later on will
be implemented during the ETL process. We only include those attributes
from the sources that are used in the SDW schema.

Table 1. Some examples of abstract specifi cation for the ETL processes.

Source fi le: District shape fi le SDW dimension: District Transformation

Source attribute SDW Attribute

CODDIST DistrictID Not required

DISTRITO PartialCode Not required

NDISTRITO Name Standardize the format

of the names considering
capitalization of only the fi rst
letter and make the necessary
corrections for accent marks.

geom. Geom Union of geometries for the

same district.