The Attribute Reduction Algorithms Research Of Rough Set Based On Concept Lattice

Inductive learning is the most important branch of machine learning. Its main purpose is to induce the general rules and patterns from data. As redudancy properties affect the performances of time and space as well as the quality of rules. Therefore, at the premise of maintaining the ability of classfication, reduction becomes the key technology of inductive learning. Because rough set theory has the advantage of anti-interference, the main existing algorithms of recudion are based on rough set theory. However, the performances of time and space as well as the completeness are not very satisfied. Because of the equivalence relationship between equivalance class of rough set and extension of concept lattice, this disseration researches the redcution of rough set data from the perspective of concept lattice. The contributions of the disseration are as follows:(1) Based on the research of the rough set theory and concept lattice theory, the equivalence theorem between equivalance class of rough set and extension of concept lattice, the determine theorem of the redundant properties, the implication relationship based on concept lattice and the potential reduction, consistency of decision table are proposed in the disseration as well as the proof of the theorems above.(2) By using the theorems proposed above, A reduction algorithm ARCL based on the concept lattice is proposed and the performance of the algorithm ARCL is detailed analysised and compared. The experiment shows that ARCL performs much better than traditional algorithms.(3) According to the above-mentioned study, the prototype system for rough set property reduction based on the concept lattice is realized, which includes three modules: data import, construction of concept lattice and reduction based on ARCL.