Exploiting comparative studies using criteria: generating knowledge from an analyst's perspective

Abstract

In this work the use of qualitative preferences for classifying and selecting MOEAs is introduced. The classical notions of the Analyst and the so called Prescriptive Analysis are introduced explicitly in EMO, identifying some difficulties in exploiting the results of the comparative studies performed by the current fashion. A methodology is developed that allows the analyst to translate DM's general preferences as well as quantitative benchmarking results into a practical tool for the comparison of MOEAs, facilitating the selection of the proper method and/or parameters for the MCDM problem at hand. A comparative experimentation is performed using well known state of the art functions, allowing drawing clear conclusions about the utility of the proposed methodology. The results are useful for research, practitioners and analysts involved in benchmarking, comparative studies and prescriptive analysis for EMO.