Analysis of Data Presented by Multisets Using a Linguistic Approach

Abstract

The problem of the analysis of datasets formed by the results of group expert assessment of objects by a certain set of features is considered. Such datasets may contain mismatched, including conflicting values of object evaluations by the analyzed features. In addition, the values of the assessments for the features can be not only point, but also interval due to the incompleteness and inaccuracy of the experts’ knowledge. Taking into account all the results of group expert assessment of objects for a certain set of features, estimated pointwise, can be carried out using the multiset toolkit. To process interval values of assessments, it is proposed to use a linguistic approach which involves the use of a linguistic scale in order to describe various strategies for evaluating objects: conservative, neutral and risky, and implement various decision-making strategies in the problems of clustering, classification, and ordering of objects. The linguistic approach to working with objects assessed by a group of experts with setting interval values of assessments has been successfully applied to the analysis of the dataset presented by competitive projects. A herewith, for the dataset under consideration, using various assessment strategies, solutions of clustering, classification, and ordering problems were obtained with the study of the influence of the chosen assessment strategy on the results of solving the corresponding problem.