Comparative Analysis of Expert Evaluation Criteria Under Z-Information

Abstract

The paper develops a method for comparative analysis of expert evaluation criteria, considering their reliability. Expert criteria are formalized on the basis of linguistic Z-numbers, the components of which are the values of linguistic variables with the properties of completeness and orthogonality. Expert criteria are a set of Z-numbers, the number of which is equal to the number of scale levels used by experts for the evaluation. For a comparative analysis of expert criteria, similarity indicators are determined based on aggregating segments of Z-numbers. To determine the aggregating segments of Z-numbers, their components are multiplied. Using the similarity indicators of expert criteria, a similarity relation is constructed, which allows one to divide the criteria into clusters of similar ones. The most numerous cluster is taken as the basis for further analysis, all elements of which are considered to be the most similar and have equal weight coefficients. Weight coefficients determine the significance of taking into account the corresponding criterion. Less numerous clusters have smaller weight coefficients, which are determined by the sums of indicators of similarity with the elements of the most numerous cluster. Depending on the values of these sums, all criteria are ranked, after which the Fishburn scale is used to determine the weight coefficients. The developed method allows analyzing expert criteria The paper develops a method for comparative analysis of expert evaluation criteria, considering the reliability of information coming from experts, filtering out unreliable and erroneous information, identifying similar criteria and determining the degree of their importance for the contribution to the group criterion.