Using Class Decomposition for Building GA with Fuzzy Rule-Based Classifiers - Developing Concepts in Applied Intelligence - page 66

Using Class Decomposition for Building GA
with Fuzzy Rule-Based Classifiers

Passent El-Kafrawy and Amr Sauber

Dept of Math and CS, Faculty of Science
Menoufia University, Egypt
{passentmk,amrmausad}@gmail.com

Abstract. A classification problem is fully partitioned into several small
problems each of which is responsible for solving a fraction of the origi-
nal problem. In this paper, a new approach using class-based partition-
ing is proposed to improve the performance of genetic-based classifiers.
Rules are defined with fuzzy genes to represent variable length rules.
We experimentally evaluate our approach on four different data sets
and demonstrate that our algorithm can improve classification rate com-
pared to normal Rule-based classification GAs [1] with non-partitioned
techniques.

Keywords: Genetic Algorithm, Rule-based Classification, Class

Decomposition, Divide and Conquer, Fuzzy Rules.

1 Introduction

The interest in decision-making has been gaining momentum in recent years.
Nowadays enormous amounts of information are collected continuously. The
tremendously growing amount of data has made manual analysis by experts
a tedious task and sometimes impossible. Many hidden and potentially useful
relationships may not be recognized by the analyst. The explosive growth of data
requires an automated way to extract useful knowledge. One of the possible ap-
proaches to this problem is by means of data mining or knowledge discovery
from databases. Rule-based classification [2,3] is one of the most studied tasks in
data mining community and is an active research area. Classification rules are
typically useful in medical problems, specially in medical diagnosis. Such rules
can be verified by medical experts and may provide better understanding of the
problem by-hand.
Numerous techniques have been applied to classification in data mining over
the past few decades, such as statistical methods, expert systems, artificial neural
networks, database systems, and evolutionary algorithms ( [4], [5], [6], [7], [8]).
Among them, genetic algorithm (GA)-based approaches have attracted much
attention and become one of the popular techniques for classification [9]. De
Jong and Spears [1] considered the application of GA to a symbolic learning
task—supervised concept learning from a set of examples. Corcoran and Sen
[2] used GA to evolve a set of classification rules with real-valued attributes.