Impute missing assessments by opinion clustering in multi-criteria group decision making problems

Multi-criteria group decision-making and evaluation (MCGDME) method typically aggregates information in evaluation tables. For various reasons, evaluation tables (decision matrix) often include missing data that highly affect correct decision-making and evaluation. Most existing imputation methods of missing data are based on statistical features which do not exist in an MCGDME setting. This paper proposes an imputation method of missing data (IMD) in evaluation tables. The IMD method measures the similarity betweent two evaluators’ mental models. Evaluators are then classed into several groups based on their similarities by using fuzzy clustering methods. Finally, missing data are imputated under the assumption that the imputated value of missing data does not change the previous clustering results. The proposed IMD method is implemented and tested in two numerical experiments.