Research Of Data Mining Based On Concept Lattice And Its Extended Models

Formal concept analysis (FCA) is a mathematical tool that describes philosophical concept: by means of formalization; it has been widely used in software engineering, information retrieval and data mining etc. Now more and more attention has been paid to the research of FCA.This thesis mainly focuses on data mining based on concept lattice and its extended models, the context was involved with two domain of FCA: generation and application of concept lattice.In generation of concept lattice, the thesis first reviews existing generation algorithms and analyzes their principles of generation, then presents a top-down algorithm based on optimal-incomplete cover. The algorithm will generate concept set and hasse diagram used width-first search of the line diagram. In addition, this thesis proposes and implements an algorithm, which generates relatively reduced concept lattice.On the other hand, concept lattice and its extended models are used in data mining. As concept extent was concerned, the thesis presents two algorithms to build the min-support lattice and the min-confidence lattice respectively and analyzes the approach of applying min-support lattice to clustering analysis in detail. As concept intent was concerned, the thesis attempts to measure distance of two concepts through relation of concept intent. As a result, the similarity-based algorithm is proposed and implemented for clustering analysis. Besides what are mentioned above, this thesis mainly lays emphasis on mining classification rules using quantitative relatively reduced concept lattice, the algorithm is more efficient than Rulearner at complexity of time and space, and eliminates the redundant rules.Moreover, this thesis compares FCA with clustering analysis, and analyzes the similarities and differences of classification based on concept lattice and decision tree.