Database Reference
In-Depth Information
tidimensional model that allows OLAP analyses.
However, classical OLAP tools are ill-adapted to
deal with complex data. OLAP facts representing
complex data indeed require appropriate tools and
aggregation methods to be analyzed. A new idea
consists in combining OLAP and data mining
algorithms to provide new OLAP operators able
to compute more significant aggregates, mainly
on complex data cubes. In this context, we have
proposed three approaches.
A
the set of distinct attribute names and
V
the set
of element and attribute values.
An XML graph can be denoted by the expres-
sion:
G
:
=
tl
ψ
, where
t
is an finite ordered tree,
l
a function that labels a node in
t
with symbols
from
E
∪
A
, and
ψ
a function that associates a node
in
t
to its corresponding value in
V
. The root node
of
t
is denoted
root
t
.
In an XML graph, a node
e
can be referenced
by another node
e
′. Let
a
and
a
′ be two attribute
nodes that are children of
e
et
e
′, respectively.
Then,
e
references
e
′ if and only if
ψ
(
a
) =
ψ
(
a
′)
and
l
(
a
)
unequals l
(
a'
). Such a link is referred to
as a
virtual key reference
.
, ,
1.
Sparse data visualization:
with multiple cor-
respondence analysis, we have reduced the
negative effect of sparsity by reorganizing
cube cells (BenMessaoud
et al.
, 2006).
xML Data Warehouse
Reference Model
2.
Complex fact aggregation:
with agglomera-
tive hierarchical clustering, we obtain aggre-
gates that are semantically richer than those
provided by traditional multidimensional
structures (BenMessaoud
et al.
, 2006a).
Previous XML data warehouse architectures
converge toward a unified model. They mostly
differ in the way dimensions are handled and in
the number of XML documents that are used to
store facts and dimensions. We may distinguish
four different families of physical architectures:
3.
Explanation of possible relationships in
multidimensional data:
we have designed
a new algorithm for guided association rule
mining in data cubes (BenMessaoud
et al.
,
2007).
1.
One XML document for storing facts and
another for storing all dimension related
information (XCube);
MODELING xML DATA
WAREHOUSES
2.
A collection of XML documents that each
embed one fact and its related dimensions
(X-Warehousing);
In this section, we mainly present our analysis of
related work and our proposal to unify and enhance
existing XML warehouse models. Our reference
model notably exploits the concept of virtual key
reference we define previously.
3.
A collection of XML documents where
facts and dimensions are each stored in one
separate document (XML-OLAP);
4.
One XML document for storing facts and
one XML document for storing each di-
mension (analogous to relational star-like
schemas).
Preliminary Definitions
An XML document is defined as a labeled graph
(an
XML graph
) whose nodes represent document
elements or attributes, and whose edges represent
the element-subelement (or parent-child) relation-
ship. Edges are labeled with element or attribute
names. Let
E
be the set of distinct element names,
A performance evaluation study of these dif-
ferent representations has been performed by
Boukraa et
al.
(2006). The authors have built four
XML mammographic warehouses with respect
Search WWH ::
Custom Search