Feature Selection in Data Envelopment Analysis: A Mathematical Optimization approach

Abstract

This paper proposes an integrative approach to feature (input and output) selection in Data Envelopment Analysis (DEA). The DEA model is enriched with zero-one decision variables modelling the selection of features, yielding a Mixed Integer Linear Programming formulation. This single-model approach can handle different objective functions as well as constraints to incorporate desirable properties from the real-world application. Our approach is illustrated on the benchmarking of electricity Distribution System Operators (DSOs). The numerical results highlight the advantages of our single-model approach provide to the user, in terms of making the choice of the number of features, as well as modeling their costs and their nature.

Mathematical optimization in classification and regression trees

“…While a black-box model could be extremely good at predicting who would benefit from a policy intervention, policy makers should be able to explain why decisions are taken, as is evident, e.g., in the COVID-19 crisis. Moreover, transparency (Chen et al 2017) is a must when, for instance, benchmarking the providers of utilities (Benítez-Peña et al 2020a) or in credit scoring in consumer lending (Baesens et al 2003), the reason being the so-called right-to-explanation in algorithmic decision-making, imposed by the European Union since 2018 (European Commission 2020; Goodman and Flaxman 2017; Wachter et al 2017). Although the term Explainable Artificial Intelligence (XAI) was coined a while ago, it is now tracking a lot of attention from different communities, see, e.g., Barredo Arrieta et al (2020), Gunning and Aha (2019), Holter et al (2018), Miller (2019).…”

“… 2017 ) is a must when, for instance, benchmarking the providers of utilities (Benítez-Peña et al. 2020a ) or in credit scoring in consumer lending (Baesens et al. 2003 ), the reason being the so-called right-to-explanation in algorithmic decision-making, imposed by the European Union since 2018 (European Commission 2020 ; Goodman and Flaxman 2017 ; Wachter et al.…”

Selection of Production Mix in the Agricultural Machinery Industry Considering Sustainability in Decision Making

“…Another study combined AHP and DEA methods to evaluate the performance of com-panies in the PV energy sector. The AHP was applied to collect expert opinions and the DEA to measure which companies are the most efficient [26], to evaluate the road safety performance of a set of European countries (or DMUs), combining the AHP and DEA method [39], and to classify organizational units, where each unit has multiple inputs and outputs [40].…”

“…The DEA method performs a global assessment of alternatives; DMUs that fall outside the efficient boundary were considered underperforming, and they needed to be further analyzed to determine the measures to improve their efficiency [40]. In this study, the results obtained in the backlash were an indication of what can be improved in each of the alternatives.…”

Variable selection in convex quantile regression: L1-norm or L0-norm regularization?

“…A vast body of literature on feature extraction for DEA has emerged. One typical choice is to use principal component analysis (PCA) to reduce the dimensionality of DEA production models (e.g., Adler & Yazhemsky, 2010;Nataraja & Johnson, 2011;Benítez-Peña et al, 2020). Another commonly used method is to utilize the multicollinearity among inputs or outputs (Wilson, 2018).…”