Attribute Reductions Of Decision Systems Based On General Relation Rough Sets And Its Generalizations By Two Ways Of Fuzzy Models

The theory of classical rough sets is based on equivalence relations, where knowledge isconsidered to be some equivalence partitions induced by equivalence relations. Data nalysis anddecision making can be made by using the classifications information hidden in equivalentclasses. Attribute reduction is an important issue in rough set theory. This paper gives a newattribute reduction model based on general binary relations and the concrete results are asfollows:1. The positive regions of decision and discernibility matritrics of attribute reductions andtheir corresponding judgment theorems have been redefined in order to generalize the existingapproaches to attribute reduction in relation decision systems.2. According to fuzzy pattern recognition theory, combining maximum of membershipdegree and fuzzy cut sets, this thesis proposes attribute reduction methods under a generalRough-Fuzzy model and a general Fuzzy-Rough model. Some fuzzy statistic examples showsthe proposed methods in this thesis are effective in reducing practical data sets.