Detecting two-dimensional projection-efficient units in data envelopment analysis under big data scenarios

In the age of big data, traditional estimation methods may struggle to process large datasets efficiently. Ali (1993) laid the foundation for improving efficiency assessment using Data Envelopment Analysis (DEA). Building on this work, we demonstrate how to detect two-dimensional projection-efficient units. This is achieved by projecting the multidimensional DEA production frontier onto two-dimensional subspaces and utilizing slope analysis to identify key efficient units. These units are then linked to their full-dimensional counterparts to define projection-efficient units. We propose using these key efficient units as a preliminary step to speed up the identification of full-dimensional efficient units or to estimate the relative density of datasets. Simulations show that our method reduces computation time for the two fastest approaches by an average of 54.2 % across different datasets.