Selected Methods of Combining Classifiers,when Predictions are Stored in Probability Vectors,in a Dispersed Decision-making System

Abstract

Issues that are related to decision making that is based on dispersed knowledge are discussed in the paper. A dispersed decision-making system that was proposed in the earlier paper of the author is used in this paper. In the system the process of combining classifiers in coalitions is very important and negotiation is applied in the clustering process. The main aim of the article is to compare the results obtained using five different methods of conflict analysis in the system. All of these methods are used when the individual classifiers generate probability vectors over decision classes. The most popular methods are considered - a sum rule, a product rule, a median rule, a maximum rule and a minimum rule. An additional aim is to compare the results obtained with using a dispersed decision-making system with the results obtained when the prediction results are aggregated directly using the conflict analysis methods. Tests, that were performed on data from the UCI repository are presented in the paper. The best methods in a particular situation are also indicated. It was found that some methods do not generate satisfactory results when there are dummy agents in a dispersed data set. That is, there are undecided agents who assign the same probability value to many different decision values. Another conclusion was that the use of a dispersed system improves the efficiency of inference.