A novel mixed binary linear DEA model for ranking decision-making units with preference information. (November 2020)

• Record Type: Journal Article
• Authors: Ebrahimi, Bohlool; Tavana, Madjid; Toloo, Mehdi; Charles, Vincent
• Is Part Of: Computers & industrial engineering. Volume 149(2020)
• Publication Date: 2020-11
• Subjects: Data envelopment analysis -- Efficient units -- Decision-makers' preferences -- Weight restrictions -- Mixed binary linear programming; Engineering -- Data processing -- Periodicals; Industrial engineering -- Periodicals; 620.00285
• DOI: 10.1016/j.cie.2020.106720
• Languages: English
• ISSNs: 0360-8352

Abstract

Highlights: We propose a new mixed binary linear DEA model by considering the decision-makers' preferences. We show assurance region type II and weight restrictions (WRs) leads to infeasibility and free production. We propose a new DEA model with assurance region type I and WRs. We prove our model eliminates infeasibility and free production problems. Abstract: Several mixed binary linear programming models have been proposed in the literature to rank decision-making units (DMUs) in data envelopment analysis (DEA). However, some of these models fail to consider the decision-makers' preferences. We propose a new mixed binary linear DEA model for finding the most efficient DMU by considering the decision-makers' preferences. The model proposed in this study is motivated by the approach introduced by Toloo and Salahi (2018). We extend their model by introducing additional assurance region type I (ARI) weight restrictions (WRs) based on the decision-makers' preferences. We show that direct addition of assurance region type II (ARII) and absolute WRs in traditional DEA models leads to infeasibility and free production problems, and we prove ARI eliminates these problems. We also show our epsilon-free model is less complicated and requires less effort to determine the best efficient unit compared with the existing epsilon-based models in the literature. We provide two real-life applications to show the applicability and exhibit the efficacy of our model.