Research On Rough Concept Lattice And Mining Method Of Classification Rule

Concept lattice, which has accurate and complete characteristics, is an effective tool for data analysis and knowledge discovery. Rough set theory is a mathematical tool of handling uncertain or inaccurate data. In order to qualify concept lattice for uncertain knowledge representation, the uncertainty of concept extent is described and the method of uncertain classification rule mining is studied by using upper and lower approximation sets of rough set theory. So, it is of profound theoretical significance and application value for improving the knowledge representation ability of concept lattice. The main research work can be summarized as follows:1. A new concept lattice structure RCL (Rough Concept Lattice) and its construction algorithm CARCL are presented. Two kinds of extents (upper approximation extent and lower approximation extent) are described by adapting upper and lower approximation sets. For a pointed intent of the rough concept lattice, its upper approximation extent is uncertain and lower approximation extent is accurate. This lattice structure reflects two relations (the certain relation and uncertain relation between objects and attributes). An example validates the correctness and efficiency of algorithm CARCL.2. A kind of classification rule with rough degree and its mining algorithm EACR based on rough concept lattice are presented. Firstly, rough degree of classification rule is defined by adapting upper approximation extent and lower approximation extent of rough concept lattice, and is used to describe the uncertainty of the classification rule. Secondly, for a pointed rough degree threshold, an algorithm EACR of classification rule based on rough concept lattice is presented. In the end, by adapting the discrete SDSS star spectra data as decision context, the experiment validates that the algorithm is correct and efficient. 3. On the basis of above, by using VC++ and Oracle9i as development tools, the system of rough concept lattice construction and classification rule extraction for SDSS star spectra data are designed and realized, and its function modules, software architecture and key technologies are elaborated. In the end, the running results show that it is feasible and valuable for construction of rough concept lattice and mining of classification rules.