Toward A Robust and Learning-Based Decision Support Model: Hybrid Preference Preservation Inverse Optimization-AHP Framework

Abstract

Industry 5.0 enhances Industry 4.0 paradigm by directing the evolution to more resilient, human- centric and sustainable ecosystem. In this context, Decision Makers(DM) require the use of more reliable and robust Decision Support Model(DSM). Hence, Multi-Criteria Decision Making(MCDM) framework offers great and consistent support. However, some existing MCDM models can contain inconsistency and need more trust enhancement.Newly, owed to the Artificial Intelligent (AI) techniques, several researchers try improving the classical MCDM frameworks by proposing hybrid AI-MCDM models. These models can identify “the best decision” based on prediction, using historical decisions Data. Nevertheless, two major limitations are listed: (i) in the case of small data set the AI prediction model are not recommended. (ii) where transparency andtrust are crucial, these models can be perceived as “black boxes” with low interpretability or explicability.To overcome these limitations, this study proposes a new collaborative framework, that incorporates a hybrid MCDM model based on “Inverse optimization (IO) and Human-In-The-Loop (HITL) validation”. The proposal is adequate for various datasetsize, gives a better explicability and improves the Decision Trust.In this paper, the improved IO-MCDM framework is applied to one of the famous MCDM method “Analytic Hierarchic Process (AHP)”. Entitled “Preference Preservation Inverse Optimization-AHP (PPIO-AHP)”. The proposed framework, in additionto enhancing the learning based on historical previous best decisions, improves AHP consistency and Preference Preservation Mechanism. Finally, a “sustainable project evaluation” is established. The result emphasizes the transformative role of” human–Inverse optimization-MCDM” collaboration, in progressing Industry 5.0’s vision of smart, sustainable, and human-centered decision making.