Synthetic Data Generation for Data Envelopment Analysis

The paper is devoted to the problem of generating artificial datasets for data envelopment analysis (DEA), which can be used testing DEA models and methods. In particular, papers that applied big often synthetic generation obtain large-scale because real large size, available in public domain, are extremely rare. This proposes algorithm takes as input some dataset complements it by efficient inefficient units. process extends part frontier inserting units, keeping original unchanged. For this purpose, uses assurance region method consistently relaxes weight restrictions during iterations. approach produces closer ones, compared other algorithms generate from scratch. proposed a pair small real-life datasets. As result, were expanded 50K Computational experiments show artificially generated DMUs preserve isotonicity do not increase collinearity whole.