PEAXAI: Probabilistic and Explainable Efficiency Analysis in R

R. González Moyano, J. Aparicio Baeza, J. L. Zofio Prieto, V. J. España Roch

PEAXAI (Probabilistic Efficiency Analysis using Explainable Artificial Intelligence) is an R package for the probabilistic estimation of technical efficiency based on machine learning (ML) and explainable artificial intelligence (XAI). PEAXAI evolves the canonical Data Envelopment Analysis method for efficiency measurement, which is deterministic and highly dependent on the observed sample, by delivering more robust and generalizable results. To this end, it departs from the DEA labeling of decision-making units (DMUs) as Pareto-efficient or inefficient and applies supervised classification techniques, including optimal balancing, to determine the probability of being efficient for the observed input-output quantities. Once the ML model is trained, PEAXAI uses XAI techniques to generate global and individual-specific insights, identify the most influential features, and provide data-driven recommendations to guide counterfactual changes aimed at achieving a given probability level.

Keywords: Data Envelopment Analysis, Machine Learning, Classification models, Counterfactual explanations, XAI

GT SW I: Paquetes de R
September 4, 2026  9:00 AM
Aula 20

I. Ortega-Fernandez, M. Sestelo, N. M. Villanueva

N. M. Romaris Lodeiro, O. Benavente Fernández, R. Fernández Casal, S. Naya Fernández, J. Tarrío Saavedra

N. M. Villanueva, M. Sestelo