Contents

List of Examples

List of Figures

List of Tables

Title and Copyright Information

Preface

• Audience
• Documentation Accessibility
• Related Documentation
• Conventions

What's New in the Oracle Data Mining APIs?

• Oracle Data Mining 11g Release 2 (11.2.0.3) API New Features
• Oracle Data Mining 11g Release 2 (11.2.0.1) API New Features

1 Data Mining API Use Cases

• Analyze Customer Demographics and Buying Patterns
  • Segment Customer Base and Predict Attrition
  • Predict Missing Incomes
  • Find Anomalies in the Customer Data
    • Find the Top 10 Outliers
    • Compute the Probability of a New Customer Being a Typical Affinity Card Member
    • Find the Demographics of a Typical Affinity Card Member
• Evaluate the Success of a Marketing Campaign
• Use Predictive Analytics to Create a Customer Profile

2 A Tour of the Data Mining APIs

• Data Mining PL/SQL Packages
  • DBMS_DATA_MINING
    • CREATE_MODEL
    • APPLY
  • DBMS_DATA_MINING_TRANSFORM
  • DBMS_PREDICTIVE_ANALYTICS
• Data Mining Data Dictionary Views
• Data Mining SQL Functions
• Data Mining Java API
  • The JDM Standard
  • Oracle Extensions to JDM
  • Principal Objects in the Oracle Data Mining Java API
    • Physical Data Set
    • Build Settings
    • Task
    • Model
    • Test Metrics
    • Apply Settings
    • Transformation Sequence

3 Creating the Case Table

• Requirements
  • Column Data Types
  • Data Sets for Data Mining
• About Attributes
  • Data Attributes and Model Attributes
  • Target Attribute
  • Numericals and Categoricals
  • Model Signature
    • ALL_MINING_MODEL_ATTRIBUTES
  • Scoping of Model Attribute Name
  • Model Details
• Nested Data
  • Nested Object Types
    • DM_NESTED_NUMERICALS
    • DM_NESTED_CATEGORICALS
  • Example: Transforming Transactional Data for Mining
  • Example: Creating a Nested Column for Mining
  • Market Basket Data
• Missing Data
  • How Oracle Data Mining Interprets Missing Data
  • Examples: Missing Values or Sparse Data?
    • Sparsity in a Sales Table
    • Missing Values in a Table of Customer Data
  • How Oracle Data Mining Treats Missing Data
  • Attribute Transformation and Missing Data Treatment

4 Preparing Text for Mining

• Oracle Text Routines for Term Extraction
• Term Extraction in the Sample Programs
• From Unstructured Data to Structured Data
• Steps in the Term Extraction Process
  • Transform a Text Column in the Build Table
  • Transform a Text Column in the Test and Apply Tables
  • Create the Index and Index Preference
  • Create the Intermediate Terms Table
    • FEATURE_PREP Calling Syntax
    • FEATURE_PREP Return Value
    • FEATURE_PREP Arguments
    • FEATURE_PREP Example
  • Create the Final Terms Table
    • FEATURE_EXPLAIN Calling Syntax
    • FEATURE_EXPLAIN Return Value
    • FEATURE_EXPLAIN Arguments
    • FEATURE_EXPLAIN Example
  • Populate a Nested Table Column
• Example: Transforming a Text Column

5 Building a Model

• Steps in Building a Model
• Model Settings
  • Specifying a Settings Table
  • Specifying the Algorithm
  • Specifying Costs
  • Specifying Prior Probabilities
  • Specifying Class Weights
  • Model Settings in the Data Dictionary
• Creating a Model
  • Mining Functions
  • Transformation List
• Model Details
• Mining Model Schema Objects
  • Mining Models in the Data Dictionary
  • Mining Model Privileges
  • Sample Mining Models

6 Scoring and Deployment

• In-Database Scoring
• What is Deployment?
• Real-Time Scoring
  • Prediction
    • Best Prediction
    • Confidence Bounds (GLM only)
    • Costs
    • Rules (Decision Tree only)
    • Probability
    • Per-Class Results
  • Clustering
    • Cluster Identifier
    • Probability
    • Per-Cluster Probabilities
  • Feature Extraction
    • Feature Identifier
    • Match Quality
    • Per-Feature Values
  • Save Scoring Results in a Table
• Cost-Sensitive Decision Making
• Batch Apply
  • Comparing APPLY and SQL Scoring Functions

7 The Data Mining Java API

• The Java Environment
• Connecting to the Data Mining Engine
  • Connection Factory
    • Create ConnectionFactory Using OraConnectionFactory
    • Lookup ConnectionFactory From the JNDI Server
  • Connect Using JDBC
  • Connect Using ConnectionSpec
  • Features of a DME Connection
    • Create Object Factories
    • Provide Access to Mining Object Metadata
    • Persistence and Retrieval of Mining Objects
    • Execute Mining Tasks
    • Retrieve DME Capabilities and Metadata
    • Retrieve Version Information
• API Design Overview
  • Describing the Mining Data
  • Build Settings
  • Enable Automated Data Preparation
  • Executing Mining Tasks
  • Creating Mining Task Workflows
  • Building a Mining Model
  • Exploring Model Details
  • Testing a Model
  • Applying a Model for Scoring Data
  • Using a Cost Matrix
  • Using Prior Probabilities
  • Embedded Transformations
    • Embed Single-Expression Transformations
    • Embed Complex Sequence of Transformations
  • Using Predictive Analytics Tasks: Predict, Explain, and Profile
  • Preparing the Data
    • Using Binning/Discretization Transformation
    • Using Normalization Transformation
    • Using Clipping Transformation
    • Using Text Transformation

Index