Title: Tuning the data sample for data envelopment analysis

Authors: Athanasios Valiakos; Vincent Charles

Addresses: Department of Informatics, University of Piraeus, 80 Karaoli & Dimitriou Str., 18534 Piraeus, Greece ' CENTRUM Católica Graduate Business School, Pontificia Universidad Católica del Perú, Lima, Peru

Abstract: Data envelopment analysis (DEA) relies on efficiency scores being relative, and, therefore, the efficiency frontier is constructed by a complete set of decision-making units. In this research, a technique is proposed using a statistical sample of large datasets, where it is proven that the efficiency frontier is not so relative since it can be calculated from a sample of the dataset. In order to assist the technique, neural networks (NNs) are also employed. Furthermore, a unified technique is proposed to acquire the efficiency scores without the use of the DEA beforehand. By obtaining a representative sample, it is easier to draw conclusions about the entire structure of the dataset with a specific error probability and accuracy. A methodology is proposed to acquire a sample based on simple random sampling technique. The DEA-NN combination is applied to the sample, while tuning the sample dataset, in order to accumulate the efficiency frontier. The NN is brought to the optimum level, producing, therefore, reliable and promising results.

Keywords: decision support systems; data envelopment analysis; DEA; large datasets; artificial neural networks; ANNs; statistical sampling; Monte Carlo simulation; operational research.

DOI: 10.1504/IJOR.2017.087280

International Journal of Operational Research, 2017 Vol.30 No.3, pp.407 - 420

Received: 15 Aug 2014
Accepted: 02 May 2015
Published online: 12 Oct 2017 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article