Database Reference
In-Depth Information
Step 2.2: Data Warehouse Maintainability using Frame
Model Metadata
The frame metadata (Fong & Huang, 1997) consists of two classes: static classes and
active. The static class stores data in its own database. It captures the semantics of heterogeneous
relational schemas after schema translation. With an object frame metadata model agent
as shown in Figure 13, frame metadata can be processed with an object-oriented view and
data operation functions. When an event occurs, it triggers a process in the constraint class,
which calls for the operations in the method class for action. Data can be actively updated
to maintain the view for decision support systems. The result is an active data warehousing
view maintenance.
To implement the web usage mining for maintaining user access patterns online, we
use frame metadata to update user access paths continuously as follows:
Header class
Class Name
Operation
Class Type
Parents
V
O
Call Insert_path
Active
Constraint class
raint class
Constraint_
Name
Method_
Name
Class_
Name
Parameter Ownership Event Sequence Timing
Insert_path
Insert_path
V
δR
Self
Insert
After
Repeat
Method class
Method class
Method_Name Class_ Name Parameter Method_
type
Condition
Action
Insert_path
V
R S , δR
Tuple
If Code =
“GET”
Insert δR into R S
Consequently, the minimum support and confi dence thresholds value must be specifi ed
by the analyst as input parameter to build the frequent tree patterns of user access paths,
which will derive the user access patterns (path traversal patterns) after data mining. Support
and Confi dence are two measures of rule interestingness. They refl ect the usefulness of
certainty of discovered rules. Each measure is associated with a threshold controlled by
users or domain experts. Rules that do not meet the threshold are considered uninteresting,
and hence are not presented to the user as knowledge. A strong association rule has a large
Support and high Confi dence level.
APPLICATIONS OF OLAM OF PATH
TRAVERSAL PATTERNS
Each query to a web usage mining system returns a set of user navigation paths/patterns.
Then the analyst faces the nontrivial problem of evaluating these patterns and deriving
reliable conclusions from them. A navigation pattern describes one or more routes among
 
Search WWH ::




Custom Search