Rough Set Extended Model And Attributes Reduction In Incomplete Information Systems

In classical Rough Set.the equivalence relations are play important role .In fact , the equivalence relations are likely to inexistence in practice,so the classical rough set theory is limited in incomplete information systems .For example,because of null value,the equivalence relations aren't constructed via attributes value. The kind of binary relation is basis for rough compute,attribute reduction and extract rules in incomplete information systems.Started with the equivalence relations of classical rough set model, we put forward four kinds of extensions rough set model by introduceding the set-pair analysis theory . Not only does the extension reserve the advantage of the present model,but also these models discard the disadvantage of the present model.Based on these,we have researched the problem of attributes reduction and an algorithm are presented for one of the these extensions models.Started with approximated set ,we introduce majority inclusion relation into incomplete information systems based on tolerance relation.This model improve the tolerance relation and make it more flexible.Furthermore,we prepose a new version of discernibility matrix ,so called variable precision discernibility matrix,which can tolerate the noise of information.In addition ,a reduction algorithm is also presented based on variable precision discernibility matrix .Started with attributes value of every obiects in universe, we extend the Rough Set model, and then apply it to the incomplete information systems in which attributions may occur as set values. Based on the co-relation between these attributes value sets, we give out tolerance relation and sim-partial order relation on such systems. Then we propose the construction method of discemable matrix based on this two kind of relations, study the approach of attribute reduction by discemable matrix in the set-valued incomplete information system, and give the judgments theorem of attributes reduction. These results are meaningful in incomplete information systems.At last, from the point of view of pansystems methodology.we study the existence of panchoas,panattractor,strange panattractor in information systems,and their relations with equivalent classses. We analysis the compatibility of decision table,and divide attributes core into pure core and impure core ,and discuss the relations between the attributes core,discernibility matrix and pansystem fixed subset. Based on all these foundations,we study the problem of attributes reduction of information systems by means of the PM, providing a new approach and idea to the processing of information systems.