Parsimoniously interactive multiple attribute rating (PIMAR) and its applications to large-scale inventory classification in the context of Industry 3.5

Abstract

This study investigates parsimoniously interactive multi-attribute rating (PIMAR) methods suitable for intelligent manufacturing in Industry 3.5, where elicitation and preservation of human intelligence are essential to enable flexible decisions. The focus is on deterministic interactive multiple criteria decision-making (MCDM) problems with many alternatives and a small number of criteria. The aim is to reduce the number of interactions with decision-makers (DMs) as much as possible, as the low- or medium-level of automation in Industry 3.5 keeps DMs busy with routine jobs, and there may be a thousand or a million alternatives in the planning system. The study proposes a research framework for a comprehensive investigation of PIMAR methods, including method design and effectiveness examinations. Various conventional MAR methods are compared with the proposed PIMAR method using synthetic artificial decision-makers (ADMs) with specific preferences generated using various weight distributions. The results show that the proposed PIMAR method outperforms conventional methods in both linear and prospect-theory utilities across various weight distributions and permutations. Furthermore, the study suggests that interactions enhance the rating effectiveness, and PIMAR methods are suitable for MCDM problems with many alternatives and a small number of criteria.