ACES: Estimating Production Frontiers for Technical Efficiency Analysis in R

V. J. España Roch, J. Aparicio Baeza, X. Barber Vallès

ACES is an R package for estimating production frontiers under standard economic shape constraints. It implements three complementary methods: ACES, which combines Multivariate Adaptive Regression Splines with monotonicity and concavity constraints; RF-ACES, an ensemble extension based on bootstrap aggregation that enhances robustness and stability; and Q-ACES, a computationally efficient heuristic version tailored for large-scale applications. These methods are proposed as a flexible alternative to traditional Data Envelopment Analysis, addressing well-known limitations such as sensitivity to dimensionality and overfitting. Beyond frontier estimation, the package provides user-friendly tools for efficiency measurement, variable selection, hyperparameter tuning, benchmarking, and simulation. Illustrative examples confirm the framework delivers more accurate and stable results than established approaches, especially in high-dimensional or structurally complex scenarios.

Keywords: Data Envelopment Analysis, Machine Learning

Data Envelopment Analysis
September 2, 2026  3:30 PM
Aula 21

J. Aparicio Baeza, M. Kapelko, L. Ortiz Henarejos, J. L. Zofío Prieto

J. F. Monge Ivars, J. L. Ruiz