In centralized organizations, senior managers typically supervise multiple comparable decision-making units (DMUs), and improving overall efficiency depends on identifying units that performance has more impact on systems. Although Data Envelopment Analysis (DEA) has been extensively applied for performance evaluation in such environments, most traditional DEA-based approaches, such as super-efficiency, focus on local efficiency instead of the collective contribution of multiple DMUs. More recent centralized DEA formulations used to identify outstanding units partially remove this limitation; however, many of these models rely on predefined constants, which may lead to infeasible or unbounded optimization problems and reduce the robustness and interpretability of the results. This study proposes a new DEA-based mixed-integer nonlinear programming model to identify a subset of k outstanding DMUs from n units operating within a centralized system. The proposed model eliminates the need for any predefined constants and guarantees feasibility and boundedness. The proposed model is validated using two empirical. From a managerial perspective, the proposed approach provides a practical decision-support tool for centralized organizations.
Cheraghali et al. (Thu,) studied this question.