The Research On Several Methods Of Attribute Reduction In Information Systems

Rough set theory, proposed by Pawlak Z. in 1982, is an efficient tool used in data analysis. The theory of concept lattice, proposed by Wille R. in the same year, is used for creating and arranging of concepts. As two powerful tools in data analysis and knowledge discovery, they have been applied in many research areas, such as data mining, machine learning, decision administering, information retrieve, and software engineering and so on.Attribute reduction is one of the key problems in knowledge discovery. In this thesis, we study attribute reduction in information systems based on rough set theory and concept lattice theory. We propose methods of attribute reduction of L fuzzy concept lattice and attribute re-duction in concept lattices based on maximal rules. The relationships of algebraic reduction and information entropy reduction are also analyzed. The main results in this thesis are summarized as follows:1. An approach to attribute reductions of L fuzzy concept lattice is proposed. The def-inition and judgment theorem of reduction based on L fuzzy formal context are given. The discernibility matrix is also constructed by employing the cut set in fuzzy mathematics. Then we can obtain all the reducts by boolean calculation. The attribute characteristics are then ana-lyzed.2. Attribute reduction in concept lattices and attribute characteristics based on maximal rules are proposed. The purpose of rules extraction is to judge a new object, and then the maximal rules are very important, the others are redundancy when they are compared with the maximal rules. The reducted decision formal context has the same condition extension and decision extension, but the reduction of condition attributes makes the judgement more convenient and the reduction of decision attributes makes the judgement more accurate. The attribute characteristics are also analyzed.3. The equivalence and entailment relations between algebraic reduction and information entropy reduction are studied in an information system. By employing a new kind of entropy, we define a definition of entropy reduction in an information system and in a decision table respectively. Then we analyze attribute reduction from the views of algebra and information entropy, and obtain the essential relation between algebra reduction and entropy reduction.