Validation of data categorization using extensions of information systems: Experiments on melanocytic skin lesion data

The purpose of data categorization is to group similar cases (items, examples, objects, etc.) together under a common label so that information can be acted upon in the aggregate form. Sometimes, this process is made arbitrary by an expert. For each case, an expert determines a class (group) to which the case is classified. In the paper, we propose a method for validation of a categorization process. The method is based on extensions of information systems defined in terms of rough sets. Usefulness of the proposed method is shown for the data used in the synthesis of images of melanocytic skin lesions.