Discernibility Measures for Fuzzy β Covering and Their Application

Area of Science:

• Information Science
• Data Mining
• Fuzzy Mathematics

Background:

• Fuzzy beta covering, a blend of fuzzy sets and covering rough sets, is a key area of research.
• The fuzzy beta neighborhood is the fundamental unit for granulation in fuzzy beta covering.

Purpose of the Study:

• To propose a novel discernibility measure for evaluating the distinguishing ability of fuzzy beta covering families.
• To introduce variants of the discernibility measure for analyzing changes in distinguishing ability.
• To formalize fuzzy beta covering decision tables and address knowledge reduction.

Main Methods:

• Introduction of parameterized fuzzy beta neighborhoods to quantify sample similarity.
• Development of joint, conditional, and mutual discernibility measures.
• Formalization of fuzzy beta covering decision tables and design of a forward attribute reduction algorithm.

Main Results:

• The proposed discernibility measures effectively evaluate the distinguishing ability of fuzzy covering families.
• The new measures share properties with Shannon entropy, offering insights into uncertainty.
• The attribute reduction algorithm effectively reduces redundant fuzzy coverings.

Conclusions:

• The developed method accurately assesses data uncertainty across various datasets.
• The proposed approach demonstrates superior performance in attribute reduction compared to existing algorithms.