Databases Reference
In-Depth Information
Some mining algorithms search for approximate dependencies that are
functional dependencies that almost hold. Such dependencies arise in many
DBs when there is a natural dependency between attributes, but some rows
contain errors and missing data or represent exceptions to the integrity
constraints. Such cases occur frequently in legacy systems. Approximate
dependencies in the legacy system can be confirmed by the user as valid
dependencies for the ongoing application.
The combination of the previous techniques makes functional depend-
ency
acquisition
possible
and
the
relational
theory
applicable
to
real
applications.
13.4.3
Mapping From Conceptual Schema to Logical Schema
Current design methodologies provide high-level conceptual models that
are either extended E/R models or object-oriented models. These models are
better suited to reducing the semantic distance between user perception and
the formal representation of that perception. However, no current DB sys-
tems support all the concepts provided in conceptual models. Consequently,
conceptual schemas are always transformed into logical models supported by
a class of DB systems. Next we discuss the transformation rules between an
extended E/R model and a relational model, and between an object-oriented
model and a relational model.
13.4.3.1
Mapping an Entity-Relationship Schema Into a Relational Schema
As a general rule, given an extended E/R model such as the one summarized
in Figure 13.4, one can say that any attribute of the conceptual level is an
attribute of the relational model, and any entity or relationship in the con-
ceptual level can be transformed as a relation in the relational model. How-
ever, conceptual attributes may be multivalued, while they are not in the
relational model. Entities and relationships may have complex attributes,
while relations have simple attributes. Conceptual schemas may contain
aggregation and generalization hierarchies, while relational schemas cannot.
Conceptual schemas have roles and cardinalities that are not supported by
the relational model. Then mappings between an extended E/R model and a
relational model are not direct mappings. There should first be some trans-
formations within the E/R model before direct transformations are applied.
We briefly summarize the two types of transformation and show how a
CASE tool can apply them with respect to a certain methodology.
