Oracle Data Miner 10.2 supercedes Oracle Data Miner 10.1.
Oracle Data Miner Release 10.2 targets the data analyst. The user interface is designed to increase the analyst's success rate in properly utilizing Oracle Data Mining algorithms in several ways:
Oracle Data Miner supports many of the new Oracle Data Mining 10.2 features. Oracle Data Mining 10.2 supports Decision Tree Models, one-class classifiers using Support Vector Machine, and a new Java interface that complies with the Java Data Mining (JDM) standard for data mining (JSR 73).
This section briefly describes new features of Oracle Data Miner 10.2:
Oracle Data Miner supports the following
A Mining Activity is a step-by-step guide for model build, test, or apply. A build activity outlines steps for data preparation, model build, and model test, where appropriate; the exact steps depend on the algorithm selected. Models can be tested either as part of a build activity or as a separate test activity. A test activity automatically performs required data preparation then model test for a model created using a build activity. To test a model that was not created using a build activity, you must perform all required data preparation. An apply activity automatically performs required data preparation and then model apply for a model created using a build activity. To apply a model that was not created using a build activity, you must perform all required data preparation.
The mining activities for Oracle Data Miner 10.2 have improved heuristics, especially for the data preparation steps.
Mining activities are now available on the Activity menu. You can select a build activity, an apply activity, or a test activity.
Names of tables and views and for mining objects may have mixed case. Such names must be enclosed in quotation marks.
Model names may not be mixed-case names.
Oracle Data Miner has revised versions of the Sample and Stratified Sample transformation wizards that are easier to use. Two new transformations have been added: Stratified Split, splits a data set into either two tables or two view while preserving the distribution of a selected attribute, and Text, which transforms a text column for use in an activity or in a PL/SQL program.
Oracle Data Miner includes model viewers for Anomaly Detection and Decision Tree models. There is also a viewer for residuals for regression models.
Oracle Data Miner supports Receiver Operator Characteristics (ROC) for classification models.
Oracle Data Miner calculates predictive confidence for classification models; for information about this metric, see Classification Model Test Metrics: Predictive Confidence. Oracle Data Miner also displays accuracy for classification models to clarify the confusion matrix, as described in Classification Model Test Metrics: Accuracy.
Oracle Data Miner provides residual plots for regression models.
Models are built using transformed data; therefore, model build and apply results are expressed in terms of the transformed data, not the original data. For example, you might get rules expressed in terms of bin numbers instead of actual values. It is easiest to use rules and other results that are expressed in terms of the original data. If you use build and apply activities, displayed results are expressed in terms of the original data.
The Predictive Analytics procedures Predict and Explain make data mining possible for a broad audience of less technically astute users by automating many of the stages of data mining. For more information, see Predictive Analytics.
You can generate PL/SQL code for a mining activity. The generated code can be saved in a file from Oracle Data Miner. The Oracle Data Miner Code Generation Extension to Oracle JDeveloper and Oracle SQL Developer generates the code and allows you to incorporate it into programs. For details, see Oracle Data Miner PL/SQL Code Generator. The detailed information includes an example of how to generate and execute PL/SQL code for an apply activity.
Earlier versions of Oracle Data Miner generated Java code. The PL/SQL packages generated by Oracle Data Miner are callable from Java.
Copyright © 2006, 2008, Oracle. All rights reserved.