Use of data envelopment analysis for clustering of decision making units

Abstract

In this paper, clustering based on data envelopment analysis (DEA) of eight educational groups, with each group having four inputs and outputs, as well as DMU_k ID, is specified.  Minitab software and optimal weights for the formation of clusters of DMU_s were used for the analysis.  The statistical descriptions of DMU_s showed, that the elementary educational group (DMU₂) with the best educational performance ranked first.  The correlation between the variables of different DMU_s was also investigated.  In an analysis of the main variables, only three items of eigenvalue were greater than one.  In clustering based on DMU_s as highlighted in green, with the four clusters, more than 97% of variables could be retained.  Based on observations as shown in red, considering the three clusters, more than 78% of information could be maintained and in this stage the abnormal variable of O₃ (scientific products) was removed from the process.  Regarding the analysis, the results of path coefficients showed that DMU₂ had seen the best economy performance.