iBARET - instance-based reasoning tool

This paper introduces a learning system iBARET (Instance-BAsed REasoning Tool). This system takes a traditional instance-based learning approach and enhances it from several points of view. It enables to solve regression as well as classification tasks. Several techniques of feature weighting combined with different methods of the model evaluation are implemented. In order to provide an effective structure of the learning system, iBARET architecture is split into two parts: Server and Interface communicating via CQL language. Attention is paid to the integrating aspects of the described tool as well. The paper discusses theoretical issues of exchange of the instance-based models mainly from the point of view of their representation and reasonable volume of this representation. PMML model supported by sophisticated reduction of the instance set is suggested for this purpose. The model is inspired by the standardized definition of the clustering