Database Reference
In-Depth Information
Chapter 12
Cost-sensitive Active and Proactive
Learning of Decision Trees
12.1 Overview
Although the inductive learning models are accurate, they suffer the
disadvantage of excessive complexity and are therefore incomprehensible
to experts pursuing high accuracy achievement, which may suce in
cases when accuracy is the most important criterion or when there is no
background or domain knowledge available. However in real world problems
it is rarely the case. In many practical problems there is an abundance of
background information that can be incorporated to achieve better and
simpler solutions. The use of background knowledge in induction processes
produces trees that are easier to understand [ Nunez (1991) ] . The use of
background knowledge is necessary when the features attached to individual
examples do not capture abstractions and general distinctions that relate
many examples of a concept. The knowledge principle is defined as “a
system exhibits intelligent understanding and action at a high level of
competence primarily because of the specific knowledge that it contains
about its domain of endeavor” [ Feigenbaum (1988) ] . The main conclusion
from this principle is that reasoning processes of an intelligent system, being
general and therefore weak, are not the source of power that leads to high
levels of competence in behavior. The knowledge principle simply says that
if a program is to perform well, it must know a great deal about the world
in which it operates. In the absence of knowledge, reasoning won't help.
One major form of background knowledge is 'cost' which needs to be
incorporated in any potential solution to make it more applicable.
In previous chapters we review a wide variety of algorithms for decision
tree induction. However, all of these algorithms assume that all errors have
183
Search WWH ::




Custom Search