COVID-19 Crisis Management: Global Appraisal using Two-Stage DEA and Ensemble Learning Algorithms

Abstract

Today, coronavirus disease 2019 (COVID-19) is a big crisis for the world. The infected countries are attempting to manage this crisis with various strategies and maximum medical capacity, but they have not been much successful. Due to the rapid growth of COVID-19 data, this paper investigated an integrated approach for performance evaluation of countries at any time of the COVID-19 pandemic. The strategies implemented in countries were summarized in three systems: prevention, infection detection, and medical. In Phase 1, after variable selection, data were collected for 100 countries with the highest infected cases by June 21, 2021. Then, mathematical modeling of two-stage data envelopment analysis with desirable-undesirable variables was performed using three basic ideas: independent, connected, and relational. By solving the relational model, the efficiency scores of the countries were obtained, and they were categorized into four classes based on these results. In Phase 2, 80% of the data were considered as training samples to generate a machine learning model via ensemble methods. In Phase 3, the class of test samples was predicted using the optimal ensemble model. The results showed that in a small dataset, the Bag algorithm had 95% accuracy in predicting the class of test samples.

An Application of Data Envelopment Analysis and Game Theory for Appraisal of Balanced Scorecard Indexes in Media Industry

“…In parallel with these problems, there have been some challenges in DEA models that researchers have answered. For example, see [6] and [7]. One of these challenges has been determining the importance of input and output variables in DEA models.…”