Abstract

Classifier systems are knowledge-based learning algorithms which take training instances as input and produce a set of rules as output Many classifier systems represent the knowledge they learn in the form of one or more decision trees. Accurate knowledgebase systems for a variety of domains have been constructed by generating decision trees using J. R. Quinlan's (1986) inductive algorithm ID3 and P. E. Utgoffs (1988) IDS. IDS is an incremental version of ID3.

Unfortunately, all these algorithms suffer from the inability to easily and effectively handle domains with numeric-valued attributes. Numeric attributes are those whose values are taken from a continuous domain or a domain with a large number of discrete values. In 1989, Sabharwal et al. extended the representational language of ID classifier systems to include a single numeric-valued attribute. The present work extends Sabharwal et al.'s w ork by developing several new algorithms for incrementally learning how to cluster numeric values. The ID decision tree construction algorithms are extended to include these new approaches. Experimental results show that these clustering algorithms improve the performance of both ID3 and IDS on domains which contain numeric attributes.

Department(s)

Computer Science

Second Department

Mathematics and Statistics

Comments

The first Author is a Graduate Student.

This report is substantially the M.S. thesis of the first author, completed May, 1991.

This thesis has been prepared in the format used by the Ablex Publishing Corporation. Pages 1-44 will be presented for publication in the book Advances in Loeic Programming and Automated Reasoning, edited by R. W. Wilkerson.

Keywords and Phrases

Automated Induction, Machine Learning, Knowledge Representation

Report Number

CSc-91-02

Document Type

Technical Report

Document Version

Final Version

File Type

text

Language(s)

English

Rights

© 1991 University of Missouri - Rolla, All rights reserved

Publication Date

1991-05-01

Download

Share

 
COinS