A reference-searching-based algorithm for large-scale data envelopment analysis computation

TL;DR

This paper introduces a novel reference-searching algorithm for large-scale data envelopment analysis that significantly reduces computation time by solving smaller linear programs, especially effective for high-dimensional data.

Contribution

The paper proposes a new DEA computation method based on reference searching, enabling efficient efficiency calculations with smaller LPs and no setup costs for subsets of data.

Findings

Computes 3 times faster than existing methods for large, high-dimensional datasets.

Flexible in computing efficiencies of data subsets without additional setup.

Can serve as a sub-procedure for other DEA algorithms.

Abstract

Data envelopment analysis (DEA) is a linear program (LP)-based method used to determine the efficiency of a decision making unit (DMU), which transforms inputs to outputs, by peer comparison. This paper presents a new computation algorithm to determine DEA efficiencies by solving small-size LPs instead of a full-size LP. The concept is based on searching the corresponding references, which is a subset of the efficient DMUs with numbers no greater than the dimension (number of inputs and outputs). The results of empirical case studies show that the proposed algorithm computes 3 times faster than the current state of the art for large-scale, high-dimension, and high-density (percentage of efficient DMUs) cases. It is flexible enough to compute the efficiencies of a subset of the full data set without setup costs, and it can also serve as a sub-procedure for other algorithms.