Concept Learning and Decision Trees

Concept learning systems often generate knowledge in the form of decision trees which are able to solve difficult problems of practical importance.

A decision tree is a representation of a decision procedure for determining the class of a given instance (Utgoff, 1989). A decision tree consists of :

• Leaf or answer nodes that indicate a classification either positive or negative.

• Non leaf or decision nodes which contain an attribute name and branches to other decision trees, one for each value of that attribute.

The top down induction of decision trees is a popular approach in which classification starts from a root node and proceeds to generate sub trees until leaf nodes are created. This approach uses attribute based descriptions and the learned concepts are represented by decision trees. It is possible to categorise conjunctive and disjunctive descriptions of concepts with 'if-then' rules which can be lifted from the trees. These rules often offer a more flexible representation than the decision trees themselves.

Table of Contents

References

Authored by Serengul Smith

E-mail to: serengul1@mdx.ac.uk
School of Computing Science Middlesex University
Revised: September 1998