Developing Membership Functions and Fuzzy Rules from Numerical Data for Decision Making

Abstract

Nowadays, decision making using fuzzy logic is a major research area for scientists, researchers and project managers. Construction of membership functions and fuzzy rules from numerical data is very important in various applications of the fuzzy set theory. Therefore, in this paper a model is proposed for development of membership functions and fuzzy rules from numerical data for decision making. The main advantage of the proposed model is its simplicity. The proposed model is applied on Fisher's Iris data for decision making. The validation result shows that proposed model has a higher accuracy than existing models.

Some Methods and Algoriths for Evaluation of Well Equipment and Regime Parameters in Rod, Depth Pump Oil Wells

“…In the simplest case, the pump should not be placed too deep or placed too shallow. Therefore, the membership function of this parameter can be selected in the form of a trapezoid or in the form of a Gaussian function [11,12]. The trapezoidal affliationfunction can be given as follows.…”

FUZZ-EQ: A data equalizer for boosting the discrimination power of fuzzy classifiers

“…(4). Besides Pedrycz, a few other authors proposed considering the histogram or the PMF/PDF of the data to construct data-meaningful fuzzy partitions [21,26,32,31]. All of them attempt to optimize the arguments of either trapezoidal or triangular membership functions to adjust fuzzy sets to the PMF/PDF of the data.…”

“…According to this statement, the definition of membership functions should depend on the empirical data distribution. To adjust the position and/or shape of the fuzzy sets, different fuzzy partitioning methods (FPMs) have been proposed [1,6,7,11,12,15,27,28,10,21,26,32,31,37]. In the context of classification, some FPMs focus on maximizing accuracy by optimizing partition parameters with respect to some criteria [6,12,27,28,10,21,26,32,31,37].…”

“…To adjust the position and/or shape of the fuzzy sets, different fuzzy partitioning methods (FPMs) have been proposed [1,6,7,11,12,15,27,28,10,21,26,32,31,37]. In the context of classification, some FPMs focus on maximizing accuracy by optimizing partition parameters with respect to some criteria [6,12,27,28,10,21,26,32,31,37]. Others prioritize interpretability over accuracy and build fixed uniform partitions with predefined parameters [7,11,15].…”

“…Besides, FPMs might differ in the stage they are applied. Some of them create the fuzzy partitions before the learning process [7,11,15,10,21,26,32,31,37], while others are part of the learning algorithm itself [1,12,27,28,6].…”