 |
|
Oracle Data Scrubbing Online System Feedback Loop
Oracle Database Tips by Donald Burleson |
Let's take a look at the all-important "feedback
loop" in the evolution of a DSS for Oracle data cleansing.
The Feedback Loop
If a suggestion was based on false assumptions
or incomplete logic, the expert will works with the Oracle DBA to
refine the model, and every iteration of the feedback loop makes the
software more intelligent and reliable.
A system that melds a human expert with a
software application that learns from mistakes is called a "Decision
Support System" (DSS). These systems use a "feedback loop" to
continuously refine the decision support software, where the human
experts explains new rules and exceptions to existing rules to the
Oracle database administrator, who incorporates these new logic into
the existing data cleansing system.
Over time, these systems get better-and better,
and a successful approach generally contains these steps:
STEP 1 - The
DSS software collect outliners (based on rules from the human
expert)
STEP 2 - The
DSS software Check and Validate each outliner
STEP 3 - The
DSS software presents the human expert with a system to manage the
"suggestions" from the expert system.
STEP 4 - The
expert approves or rejects each suggestion:
·
Approves "correct" changes (starting an audit) of the
change
·
OR
·
Rejects invalid suggestions, stating the reason
(usually an "exception to the rule"), and notifying the DBA to
improve the software by incorporating the new decision rule
exception.
STEP 5 - We
then return to STEP 1, and repeat until:
·
We have completely and reliable replicated the
decision rules of the human expert
·
OR
·
Complex human intuition prohibits further refinement
of the software.
Graphically, the iterative improvements of
Oracle expert systems for data cleansing might look like this
(Figure 1):

For expert Oracle data cleansing support and
data scrubbing consulting, use an expert from BC. We
understand the powerful Oracle data unification tools, and
we can aid in improving the data quality of any Oracle
database, large or small. |
 |
|