Decision Information System Attribute Reduction Algorithm Based On Rough Set Theory

After almost thirty years of development, Rough Set Theory has been widely accepted and adapted as an active part in the study of intelligent information science. As a mathematical tool, Rough Set Theory can fairly effectively process those incomplete, fuzzy and uncertain statistics merely with the canned data in the knowledge information system, and does not need any cut-and-dried or hypothetical information.Attribute reduction and solving attribute core occupy the key point in Rough Set Theory. One of the development directions of Rough Set Theory is the effective, fast attribute reduction of decision table, especially concentrating on attribute reduction and solving attribute core of inconsistent decision tables. This article focuses on studying and proposes two kinds of attribute reduction algorithm of solving the inconsistent decision tables in practical applications.This article firstly proposes a kind of equivalence class-based relative attribute reduction algorithm in inconsistent decision information system. This kind of algorithm points out the primary cause of the limitations when coping with the incompatible issues is mistakenly restricting in a single attribute object and not putting the equivalence class U/IND(C) as a whole. Through improving the discernibility matrix, it solves attribute reduction of completely inconsistent decision table and general inconsistent decision table. Furthermore, through establishing relative discernibility matrix, it simplifies the process of logical operations, reduces the computation time.Then this article points out a kind of conversion decision information-based attribute reduction algorithm. Through increasing attribute column which can distinguish the inconsistent objects, this kind of attribute reduction converses the complex incompatible decision table to simple compatible decision table.Finally, the comparative experiment on the UCI database can prove that the two algorithms in this article are fast, effective and have utility value.