Chapter 6. Case study software components 167
An administrator has access to the full functionality of Query Patroller Center.
The following list shows some of the tasks that administrators can do with Query
Patroller Center:
򐂰 Manage the Query Patroller system parameters.
򐂰 Create, update, or delete profiles for Query Patroller submitters and
operators.
򐂰 Create, update, or delete submission preferences for Query Patroller
submitters.
򐂰 Create, update, or delete query classes.
򐂰 Monitor and manage queries that have been intercepted by the Query
Patroller system.
򐂰 Generate and analyze reports that display database usage history.
A submitter has access to a subset of the functionality of Query Patroller Center.
The following list shows some of the tasks that submitters can do with Query
Patroller Center:
򐂰 Monitor and manage queries that they have submitted through the Query
Patroller system.
򐂰 Store results of the queries that they have submitted for future retrieval.
򐂰 Show or file results of the queries that they have submitted.
򐂰 Create, update, or delete their own query submission preferences.
Query Patroller command line support
Command line support enables Query Patroller administrators and submitters to
perform most Query Patroller tasks from the DB2 CLP or from the operating
system's command line prompt. Query Patroller commands can also be
combined with shell scripts or languages such as Perl, awk, and REXX.
6.1.7 DWE Mining
Data Mining is embedded into DB2 through the use of DB2 stored procedures
and user-defined functions (UDFs). That are three distinct modules:
򐂰 Modeling
򐂰 Scoring
򐂰 Visualization
A typical data modeling process is based on the steps depicted in Table 6-2 on
page 168. As a prerequisite, data should have been preprocessed.
168 Improving Business Performance Insight
Table 6-2 Common mining steps
The data mining run is accomplished by:
򐂰 DWE Mining Editor
򐂰 Easy Mining Procedures
򐂰 SQL Stored Procedures
Through the use of the DWE Mining Editor, you can compose mining tasks as an
integrated part of end-to-end data flows using a graphical canvas.
The Easy Mining interface is a high level API composed of Java UDFs with a
simplified interface to do a mining run with a single call to an SQL stored
procedure.
SQL Stored Procedures with SQL/MM API is a detailed expert API composed of
UDFs for the mining tasks, models, and test results. UDMs are used for defining
data mining tasks and stored procedures to build and test models. UDFs, and
table UDFs, are used for analyzing built models such as model signature, model
statistics, and model quality.
An example of the use of these features is depicted in Figure 6-31 on page 169.
Step Module Description
1 Modeling Defining a mining task
2 Modeling Doing a mining run and building a model
3 Visualization Visualizing the model
4 Scoring Scoring new data against the model
(prediction)
Chapter 6. Case study software components 169
Figure 6-31 Data mining run
The visualization functionality uses Java visualizer for Predictive Model Markup
Language (PMML) for full functionality with models created by DWE V9.1 and
IBM DWE Mining. PMML models from other providers can be visualized.
Currently, the PMML model versions supported are V2.1, V3.0, and V3.1.
You can read and modify models stored as files in PMML format and in DB2 as
objects of the SQL/MM types. Visualization is available as a stand-alone
application as part of DWE V9.1 or a Java Applet on an HTTP Server. A
Visualization sample can be seen in Figure 6-32 on page 170.
DB2
Data
BldTask
MiningData
Settings
DataSpec
Model
Modeling
SQL
Stored Procedures
Easy Mining
Procedures
DWE Mining
Editor
170 Improving Business Performance Insight
Figure 6-32 Visualizing a model
The Scoring functionality is accomplished by the use of the Easy Mining
Interface, which is based on Java UDFs to do a scoring run with a single call to
an SQL stored procedure and store the scoring result in a view.
Scoring can also be accomplished with SQL/MM APIs by the use of UDTs for the
models, result specs, and scoring results and UDFs to import models into DB2,
score data against models, and analyze the scoring result.
An example of Scoring is depicted in Figure 6-33 on page 171.
DB2
Model

Get Improving Business Performance Insight . . . with Business Intelligence and Business Process Management now with O’Reilly online learning.

O’Reilly members experience live online training, plus books, videos, and digital content from 200+ publishers.