Page 207 - DCAP603_DATAWARE_HOUSING_AND_DATAMINING

Unit 10: Multi-dimensional Data Models and Aggregation

Many conceptual data models exist with different features and expressive powers, mainly notes
depending on the application domain for which they are conceived. As we have said in the
Introduction, in the context of data warehousing it was soon realized that traditional conceptual
models for database modelling, such as the Entity-Relationship model, do not provide a suitable
means to describe the fundamental aspects of such applications. The crucial point is that in designing
a data warehouse, there is the need to represent explicitly certain important characteristics of the
information contained therein, which are not related to the abstract representation of real world
concepts, but rather to the final goal of the data warehouse: supporting data analysis oriented
to decision making. More specifically, it is widely recognized that there are at least two specific
notions that any conceptual data model for data warehousing should include in some form:
the fact (or its usual representation, the data cube) and the dime on. A fact is an entity of an
application that is the subject of decision-oriented analysis and is usually represented graphically
by means of a data cube. A dimension corresponds to a perspective under which facts can be
fruitfully analyzed. Thus, for instance, in a retail business, a fact is a sale and possible dimensions
are the location of the sale, the type of product sold, and the time of the sale.
Practitioners usually tend to model these notions using structures that refer to the practical
implementation of the application. Indeed, a widespread notation used in this context is the
“star schema” (and variants thereof) in which facts and dimensions are simply relational tables
connected in a specific way. An example is given in Figure 10.14. Clearly, this low level point of
view barely captures the essential aspects of the application. Conversely, in a conceptual model
these concepts would be represented in abstract terms which is fundamental for concentration on
the basic, multidimensional aspects that can be employed in data analysis, as opposed to getting
distracted by the implementation details.

figure 10.14: an example of star schema

Before tackling in more detail the characteristics of conceptual models for multidimensional
applications, it is worth making two general observations. First, we note that in contrast to
other application domains, in this context not only at the physical (and logical) but also at

LoveLy professionaL university 201