Hybridization of Decision Trees with other Techniques - Data Mining with Decision Trees: Theory and Applications - page 237

Chapter 15

Hybridization of Decision Trees
with other Techniques

15.1 Introduction

Hybridization in artificial intelligence (AI) involves simultaneously using
two or more intelligent techniques in order to handle real world complex
problems, involving imprecision, uncertainty and vagueness. Hybridization
is frequently practiced in machine learning, to make more powerful and
reliable classifiers.
The combination or integration of additional methodologies can be
done in any form: by modularly integrating two or more intelligent
methodologies, which maintains the identity of each methodology; by
fusing one methodology into another; or by transforming the knowledge
representation in one methodology into another form of representation
characteristic to another methodology.
Hybridization of decision trees with other AI techniques can be
performed by using either a decision tree to partition the instance space
for other induction techniques or other AI techniques for obtaining a better
decision tree.

15.2 A Framework for Instance-Space Decomposition

In the first approach, termed instance-space decomposition (ISD) and
involving hybrid decision tree with other inducers, the instance space of
the original problem is partitioned into several sub-spaces using a decision
tree with a distinct classifier assigned to each sub-space. Subsequently,
an unlabeled, previously unseen instance is classified by employing the
classifier that was assigned to the sub-space to which the instance belongs.

237