Discretization Of Continuous Attributes In Information Systems And Rules Extraction

The base of rough set theory is set theory. It can only deal with discrete data. So continuous attributes must be discreted. Researching the method of discretiz- ing attribute has important signification in theory and realistic.In this paper we have studied the methods of the discretization of continuous attributes and obtain knowledge from incomplete information systems.1.The fuzzy clustering methods of the discretization is based on transitive closure.It has passed through a series of identical transformation. It may lose a lot of information in discretizated decision-making.We have gaven a method of the discretization of fuzzy clustering methods based on perturbation(FCMBP).2.Mass of methods of the discretization adopt knowledge of domain,and it can only be applicable in It's domain.We think over the importance of attributes and optimize result of discretization with self-organizing feature map(SOFM).We gave a method of the discretization of rough set's continuous attribute based on the self-organizing feature map.3.Mass of methods of the discretization have not think over relativity between before discretization and discretizated.The discretization of attributes are accomplished independence.It can change indiscernibility relation in information system and induce illogicality and redundance dot.We gave discretization of continuous valued attributes based on interdependence of information system.4.The rough set theory, as a soft compute method, is a new mathematical tool to deal with vagueness and uncertainty. It can obtain knowledge from data objectivity.There are a lot of imperfection information system in fact. Distribution reduction, maximum distribution reduction and assignment reduction are introduced into the incomplete information systems. On this condition we presents algorithms for assignment reduction, maximum distribution reduction.Algorithms for reduction are validated by experiments. It shows that...