Hierarchical Rough Classifiers

The major applications of rough set theory in data mining are related to the modeling of concepts using rough classifiers, i.e., the algorithms classifying unseen objects into lower or upper approximations of concepts. This paper investigates a class of compound classifiers called multi-level (or hierarchical) rough classifiers (MLRC). We present the most recent issues on the construction of such classifiers from data using concept ontology as an additional domain knowledge. The idea is based on the bottom-up manner to gradually synthesize the multi-layer rough classifier for the complex target concept from the simpler classifiers. We illustrate the proposed method by experiments on real-life data.