Research On Discovery Of Classification Rules Based On Concept Lattice

Data Mining has been an urgent need because of increasing size of current databases. Being of explicitness, simpleness and completeness, Concept Lattice has been the one focus of researchers on AI, but low performance in construction and rules induction come along with the structure .In this thesis, we consider improving the performance by simplifying the structure from the angle of classification. Our research is about the classification problems on data with and without class labels attribute oClassification with class label is mainly focus on dealing with noise, reconstruction of Concept Lattice, simplification of classification rules and a classification algorithm on class labeled data has been implemented. For classification without class label, we discuss how to find a class attribute as a class label, and on this basis we then introduce support_coefficient to construct approximate Galois structure.,and extraction of classification rules has also been discussed.