Affiliations: SAIC 134 Holiday Court, Suite 318 Annapolis, MD 21401,
USA | Department of Naval Architecture and Marine
Engineering, University of Michigan, Ann Arbor, MI 48109, USA. Tel.: +1 734
763 3081; Fax: +1 734 936 8820; E-mail: [email protected]
Note: [] Formerly Research Fellow, University of Michigan, Ann Arbor, MI
48109, USA
Abstract: This paper investigates the applicability and feasibility of
real-coded GAs for the complete development of fuzzy systems with imbedded
constraints; i.e., automatic design of both fuzzy membership functions and
rules subject to certain constraints. In this paper, a real-coded GA called RGA
is developed, in which chromosomes are represented as real vectors. The
constraints associated with the fuzzy systems are explicitly considered and
satisfied during the training process of the fuzzy systems by the RGA. The
satisfaction of the constraints is done primarily by using tailor-made genetic
operators, which are designed based on the coding structure and the required
characteristics of the fuzzy systems. We apply the RGA to the design of three
fuzzy systems, called fuzzy decision models (FDMs), to model and forecast
economic activities in the crude oil tanker sector of maritime transportation.
The effects of the RGA's control parameters on the performance of the RGA,
including the crossover and mutation rates, multiple mutation, and selection
pressure, are investigated in this paper. Our results indicate that the RGA is
robust in developing appropriate FDMs with strong modeling and forecasting
capability.
