Rough Set Approach to Approximation of Concepts from Taxonomy

Abstract. We present a hierarchical learning approach to approxima-tion of complex concept from experimental data using concept taxonomy as a given domain knowledge. The proposition is based on rough set and rough mereology theory. We examine the eectiveness of the proposed approach by comparing it with standard learning approaches with re-spect to dierent criteria. Our experiments are performed on benchmark data set as well as on articial data sets generated by a road trac simulator.