REDUÇÃO DE ATRIBUTOS E GERAÇÃO DE REGRAS PARA CLASSIFICAÇÃO UTILIZANDO A TEORIA DOS ROUGH SETS

Abstract

The Rough Sets Theory (RS) [3] was proposed as
a mathematical model for knowledge representation,
uncertainty handling and sets that can not be characterized
precisely based on it's available attributes. Refers to the
indiscernibility that comes when you can not distinguish
elements of a set and they all seem to be a single element. It does not need information about data as basic probability
assignment (Dempster-Shafer’s theory) and degrees of
relevance (Fuzzy Theory). These concepts are useful when
applied in Knowledge Discovery in Database. The project is
in development, the experiments are performed with the
Rosetta tool, providing a basis for the reduction of RS
attributes through strongholds. The reduced set returns to the RS generating decision rules. The credibility of these rules is associated with an element of credibility by a function of relevance, assessing their performance. The partial results confirm RS as a good option.

Index Terms – Aproximação Inferior, Aproximação Superior, Indiscernibilidade, Região de Fronteira, Teoria dos Rough Sets.