Distributed relational db inductive learning algorithm learning algorithm "DRILA"

Dissertant

al-Ibrahim, Ali Muhammad

Thesis advisor

Abu Saud, Salih Muhammad

Comitee Members

Kanan, Ghassan Jaddu
Shalabi, Riyad
Abu Salamah, Walid

University

Arab Academy for Financial and Banking Sciences

Faculty

The Faculty of Information Systems and Technology

Department

Department of Management Information Systems

University Country

Jordan

Degree

Ph.D.

Degree Date

2008

English Abstract

Most of the data analysis approaches typically assume the input data stored in single table (i.e.

ila, ila2; ila-2000); they cannot analyse relational data without first transforming it into a single table.

This transformation, however, is not always easy and results in the lost of the structural information that could potentially be useful for the data mining processes.

Relational inductive learning research aims to develop data analysis solutions for distributed relational data without requiring it to be transformed into a single table.

This thesis presents an investigation into distributed relational inductive learning for building rule-based classifiers to predict classes of previously unseen objects which are stored in and managed by a relational database management system.

In this thesis a Distributed Relational Inductive Learning Algorithm (DRILA) has been developing by adapting the ila2 propositional rule induction algorithm for learning classifiers from relational data.

Using the new relational learning algorithm, the RILA relational learning system was developed with two rule selection strategies; the select early strategy and the select late strategy.

The select late strategy requires more learning time than the select early strategy but is more effective in finding the most efficient rule sets.

In (DRILA) the select strategy, rule selection is performed after the hypothesis search process is completed.

It is similar to the rule selection strategies used in well-known relational rule induction algorithms such as the WARMR algorithm.

Three different pruning heuristics were used to control the number of hypotheses generated during the learning processes.

Experimental results are presented on the two data sets ; KDD Cup 2001 genes data set and mutagenesis data set.

Unlike many other relational learning algorithms, the DRILA algorithm does not need its own copy of distributed relational data to process it.

This is important in terms of the scalability and usability of the distributed relational data mining solution that has been developed.

The architecture proposed can be used as a framework to upgrade other propositional learning algorithms to relational learning.

Main Subjects

Information Technology and Computer Science

Topics

• Management
• Data processing
• Computer software
• Computer networks
• Databases
• Relational databases

No. of Pages

119

Table of Contents

• APA
• MLA
• AMA

Language

English

Data Type

Arab Theses

Record ID

BIM-306394