Dispersed decision-making system with fusion methods from the rank level and the measurement level - A comparative study

Abstract This article discusses the problem of decision-making based on dispersed knowledge that is stored in several independent knowledge bases. The dispersed decision-making system, which was proposed in a previous paper of the authors, is used. In this study, four fusion methods from the rank level and nine methods from the measurement level were used in this dispersed system. These methods were tested on three data sets from the UCI Repository – Soybean, Vehicle Silhouettes and Landsat Satellite. The sets are diverse in terms of the number of objects, the number of conditional attributes and the number of decision classes. There are also various types of conditional attributes in these sets. The experimental section is divided according to the three objectives of the article. The fusion methods were compared in the two groups – rank and measurement levels. In addition, experiments were carried out fusing multiple methods simultaneously in the decision-making process. Methods from the rank level and the measurement level were applied simultaneously in the same decision-making process. Then the decisions that were generated by the methods were merged. The results were compared and conclusions were drawn. The third goal of the article was to compare the efficiency of the inference of fusion method with and without the use of a dispersed system. It was found that the use of a dispersed system improved the efficiency of inference in most cases.