Improving Expressive Power in Modeling Data Warehouse and OLAP Applications - Evolving Application Domains of Data Warehousing and Mining - page 37

Figure 17. Different options for transformation of the fact relationship from Figure 16

Notice that at the logical relational or OLAP
implementation level there is no difference
between non-strict hierarchies and multivalued
dimensions. Therefore, in our opinion, it is not
necessary to introduce these kinds of dimensions,
since they can be transformed and implemented as
non-strict hierarchies. The better choice would be
to propose and explain the required transforma-
tion in order to avoid the existence of multivalued
dimensions.

query DW data using dimensions for expanding
the perspectives of analysis and hierarchies for
aggregating measures through the drill-down and
roll-up operations.
The ongoing practice in modeling data for
DW and OLAP applications is to use a relational
model and develop star or snowflake schemas.
In addition, many OLAP systems also include an
interface that allows the designer to represent dif-
ferent elements of the multidimensional schema in
a more abstract way. However, neither relational
tables nor OLAP interfaces are able to clearly
represent different kinds of hierarchies, e.g.,
unbalanced or non-covering, and different kinds
of dimensions, e.g., role-playing or degenerate,
that exist in real-world applications. Therefore,
users cannot express clearly their analysis needs
and developers cannot implement them.
In this chapter we have chosen a simple analysis

concluSIon

Data Warehouses (DWs) are defined using a
multidimensional view of data, which is based
on the concepts of facts, measures, dimensions,
and hierarchies. On-line Analytical Processing
(OLAP) systems allow users to interactively

Figure 18. Many-to-many relationship for representing multivalued dimensions