Fuzzy interpretation of discretized intervals

When there are both numerical and nominal attributes in a database, existin g data mining systems (such as rule induction and decision tree constructio n) discretize numerical domains into intervals and the discretized interval s are treated in a similar way to nominal values during induction. This cor respondence describes a type of fuzzy intervals implemented in the HCV (Ver sion 2.0) rule induction software (referred to as HCV hereafter) for the in terpretation of rule induction results when rules with sharp intervals do n ot clearly apply to a test example at hand. A battery of experimental resul ts with HCV show that these fuzzy intervals are useful.