COMPARISON OF TWO METHODS FOR GENERATING THE COALITIONS OF CLASSIFIERS AND TWO METHODS FOR REDUCING DIMENSIONALITY IN A DISPERSED DECISION-MAKING SYSTEM

Abstract

In this paper, we consider a system in which knowledge in a dispersed form is available. In the system local classifiers are combined into coalitions. Two methods of combining classifiers in coalitions are discussed in this paper — with a hierarchical agglomeration algorithm and with Pawlak’s conflict model. The purpose of this paper is to apply methods for reducing dimensionality in these two approaches. Two methods of attribute reduction are considered — based on the rough set theory and based on attribute correlation with decision class. The most important conclusions formulated in the paper are as follows. The use of attribute selection method improves the quality of classification of the dispersed system. Better results are generated by the system with a hierarchical agglomeration algorithm.