A data mining approach for fuzzy classification rule generation
Dianhui Wang, Tharam Singh Dillon, Elizabeth Chang · 2002
This paper aims at developing a data mining approach for fuzzy classification rule generation. A regularization theory based theoretical framework for refining fuzzy classification rules is proposed. Our fuzzy rule induction methodology has four phases, namely: (1) ellipsoidal crisp rule generation with membership function assignment, (2) generic hyperbox fuzzy rule (GHR) derivation, (3) refinement of the GHR using a regularization model, and (4) simplification of the GHR by selecting an informative subset of premises out of the initial set.