The Properties Generality Principle and Knowledge Discovery Classification

Abstract

The paper examines the actual problem of automatic detection of hidden interpretable patterns in intelligent systems. The conceptual basis of the process of learning from examples is determined by the methods of class description and separation. Three basic principles are known: enumeration of class members, generality of properties and clustering. We propose an original method for implementing the principle of generality of properties based on the search for combinations of features that provide class distinction. The eff ectiveness of the approach is confi rmed by the results of numerical experiment.