Date of Award

12-2010

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Legacy Department

Computer Science

Advisor

Stevenson, Dennis E

Committee Member

Page , Edward W

Committee Member

Madison , Alan W

Committee Member

Cox , Christopher L

Abstract

Learning classifier systems (LCS) have been successful in generating rules for solving classification problems in data mining. The rules are of the form IF condition THEN action. The condition encodes the features of the input space and the action encodes the class label. What is lacking in those systems is the ability to express each feature using a function that is appropriate for that feature. The genetic algorithm is capable of doing this but cannot because only one type of membership function
is provided. Thus, the genetic algorithm learns only the shape and placement of the membership function, and in some cases, the number of partitions generated by this function. The research conducted in this study employs a learning classifier system to generate the rules for solving classification problems, but also incorporates multiple types of membership functions, allowing the genetic algorithm to choose an appropriate one for each feature of the input space and determine the number of partitions generated by each function. In addition, three membership functions were introduced. This paper describes the framework and implementation of this modified learning classifier system (M-LCS). Using the M-LCS model, classifiers were simulated for two benchmark classification problems and two additional real-world
problems. The results of these four simulations indicate that the M-LCS model provides an alternative approach to designing a learning classifier system. The following contributions are made to the field of computing: 1) a framework for developing a learning classifier system that employs multiple types of membership functions, 2) a model, M-LCS, that was developed from the framework, and 3) the addition of three membership functions that have not been used in the design of learning classifier
systems.

Share

COinS