An inexact reasoning technique using linguistic rule matrix transformations

Abstract

A technique for dealing with inexact reasoning using linguistic rule matrix transformations is presented. It is assumed that the linguistic terms are represented by trapezoidal fuzzy numbers. However, the linguistic terms can also be represented by other membership functions, such as the S-function, the pi function, and the 'close to' function. In this case, the fuzzy numbers 'plausible', 'probable', 'likely', 'pretty true', 'sort of true', 'rather true', 'close to 0.8', 'more or less close to 0.8', etc. also can be used to represent the linguistic truth values of the conditions and the certainty factors of the rules.<>