Clustering public hospitals based on crisp and fuzzy clustering techniques and probabilistic fuzzy efficiency estimates

Practical applications of data envelopment analysis (DEA) present several procedures including homogeneity of units and are assumed to be undertaking and producing similar activities. Previous papers creating comparable groups have been limited to generating isotonic groups by using various clustering techniques since there is considerable room exists. In this paper, homogenous hospital groups are created by using crisp and fuzzy grouping techniques. K-means, fuzzy c-means, and self-organizing map clustering techniques were applied by changing the hyperparameters of these techniques. Then, a fuzzy possibility DEA approach is applied to explore which hospitals are efficient, and grounded on primal and dual models. The results identify that there are five hospitals in the best hospital group and a teaching university hospital, which is located in the southeast part of the country, is efficient according to Lertworasirikul et al.'s (2003) fuzzy DEA model and a reference hospital for others. This study also highlighted several insights interrelated with strategic grouping and fuzzy efficiency estimation. Primal efficiency results are increasing while changing the α parameter from 0 to 1. This study also focused on various insights about particular pitfalls of DEA such as the creation of homogenous groups and fuzzy efficiency estimation.