Reimagining multi-criterion decision making by data-driven methods based on machine learning: A literature review

Abstract

Multi-criterion decision making (MCDM) methods can derive alternative rankings as solutions to decisionmaking problems based on survey or historical data about the performance or preference information of alternatives regarding multiple criteria. Today’s information age makes it easy to accumulate data, but also brings challenges to MCDM, such as massive data and weak data correlation. The real data in the information age should be fully utilized to put MCDM from theoretical formulation into practical application. In this regard, machine learning technologies that can adaptively discover rules as well as patterns from data of different types and characteristics show great application potential. This study dedicates to exploring the status of implementing machine learning technologies in solving MCDM problems. The related work and research advances of MCDM and machine learning technologies are briefly described. A bibliometric analysis is conducted to provide an overview of the research status, hotspots, and trends. Then, we summarize the challenges of implementing MCDM in the information age in four aspects, around which we review the research status of applying machine learning technologies to criteria extraction, criteria interaction, parameter determination, and integrated solutions of MCDM. Also, the fields of practical applications about the subject matter including business management, industrial engineering, sustainable development, emergency management, along with other fields are reviewed. This study outlines how machine learning technologies contribute to MCDM. The lessons learned from the review and future research directions are discussed. It is hoped that this review can serve as a reference and provide convenience for scholars and practitioners in the fields of decision analysis and machine learning.