An easy-to-implement fuzzy expert package with applications using existing Java classes

摘要

In this study, we present a computer package to implement fuzzy expert systems using Java. The purpose of this study is to provide an easy-to-understand and easy-to-implement computer package for those people who want to develop their own customized Java-based fuzzy expert applications without using existing end-user fuzzy expert software. The code size of the proposed package is less than 300 code lines where some existing Java classes developed by Sun Microsystems, Inc., are utilized to handle necessary geometrical manipulations. We demonstrate the ideas of how we implement fuzzy expert systems with the existing Java classes by providing code fragments. To validate the proposed package, we first developed a console application to compare the results of using the proposed package with those of using an existing commercial fuzzy expert software package. Fuzzy Inference Development Environment (FIDE). The proposed package is well-modularized in software modules with high maintainability and customizability. To validate this, we developed three additional applications with the proposed package including a window-based application and two web-based applications: JSP Model 1 and JSP Model 2 (MVC design pattern). Through the developments of these applications, we find that it is extremely simple and straightforward to build customized fuzzy expert applications with the code fragments provided in this study, saving a lot of programming works. In additions, the proposed package has good extensibilities to establish piecewise linear membership functions and to approximate non-linear membership functions. Meanwhile, UML is also utilized to describe the structures and behaviors of the proposed package.