Multi-Relational Data Mining: An Introduction

1
Multi-Relational Data Mining An Introduction

• Joe Paulowskey

2
Overview

• Introduction to Data Mining
• Relational
• Data
• Patterns
• Inductive Logic Programming (ILP)
• Relational Association Rules
• Relational Decision Trees
• Relation Distance-Based Approaches

3
Relation Data

• Relational Database
• Multiple Tables
• Defined
• Views
• Tables

4
Relational Pattern

• Multiple Relations from a relational database
• More Expressive
• Opens up
• Classification
• Association
• Regression

5
Relational Pattern (Cont.)

• Expressed in Subsets of First Order Logic

6
Data Mining

• Look for patterns in data
• What do you discover?
• Associations
• Sequences
• Classifications
• Goals of Data Mining
• Predict
• Identify
• Classify
• Optimize

• Uses
• Business Data
• Environmental/Traffic Engineering
• Web Mining
• Drug Design

7
Data Mining Relational Databases

• Most Data Mining approaches deal with single
  tables
• Not safe to merge multiple tables into one single
  table
• Number of patterns increases
• Explicit constraints required

8
Inductive Logic Programming (ILP)

• Logic Programs used to find patterns
• Clauses
• Head and Body
• Literals
• Types
• Definite
• Program

9
ILP (Cont)

• Predicate
• Relations in relational database
• Arguments -gt Attributes
• Attributes are Typed
• Database Clauses are typed program clauses
• Deductive Database

10
Relational Rule Induction ILP

• Learn logical definitions of relations
• Classification
• Rules can be found by decision trees
• Simple Algorithm
• Dealing with noisy/incomplete data

11
ILP Problems to Propositional Forms

• Propositional
• attribute-value
• Use Single Table Data Mining algorithms
• LINUS
• Background Knowledge

12
ILP/RDM Algorithms

• Share
• Learning as a Search Paradigm
• Differences
• Representation of Data, Patterns
• Refinement operators
• Testing Coverage
• Upgrading from Propositional to Relational

13
Relational Association Rules

• Frequent Patterns
• Determining Frequency
• Itemsets
• Association Rules
• Obtained by frequent itemsets

14
Relational Decision Trees

• Used for Prediction
• Binary Trees
• First Order Decision List

15
Relational Distance-Based Approaches

• Calculated distance between two objects
• Statistical Approaches

16
Conclusion