A flexible self-learning model based on granular computing

Granular Computing(GrC) is an emerging theory which simulates the process of human brain understanding and solving problems. Rough set theory is a tool for dealing with uncertainty and vagueness aspects of knowledge model. SMLGrC algorithm introduces GrC to classical rough set algorithms, and makes the length of the rules relatively short, but it can not process mass data sets. In order to solve this problem, based on the analysis of the hierarchical granular model of information table, the method of Granular Distribution List(GDL) is introduced to generate granule, and a granular computing algorithm(SLMGrC) is improved. Sample Covered Factor(SCF) is also introduced to control the generation of rules when the algorithm generates conflicting rules. The improved algorithm can process mass data sets directly without influencing the validity of SLMGrC. Experiments demonstrated the validity and flexibility of our method.