Mining Classification Rules from Datasets with Large Number of Many-Valued Attributes