Parallel Processing and Large-Scale Datasets in Data Envelopment Analysis

In order to measure the performance evaluation of a set of decision-making units (DMUs), a general data envelopment analysis (DEA) model should be solved once for each DMU. In data enabled analytics, when a large-scale dataset is evaluated, the elapsed time to apply a DEA model substantially increases. Parallel processing allows splitting the task into several parts so each part can simultaneously be executed on different processors. This study explores the impact of parallel processing to apply a DEA model for a large-scale dataset. The existing methods are clearly explained including their pros and cons. The methods are compared on different datasets according to three parameters: cardinality, dimension, and density. The strength of each existing method is changed when cardinality, dimension, density, and the number of processors in parallel are changed. A new methodology is proposed using the combination of two existing methods. In general, the proposed method is faster than all existing methods regardless of cardinalities, dimensions, and densities.