The maximum directional distance function (MDDF) model has been utilized extensively in diverse fields. However, due to practical constraints such as geographical factors, infrastructure limitations, and network scale, the theoretically improved projection paths derived from this model are often infeasible in real‐world applications. To address this limitation, this study proposes an improved MDDF model that can generate more feasible improved projection paths for decision‐making units (DMUs). First, five widely adopted clustering methods are applied to group the DMUs based on their similarities, which effectively reduces the constraints during the improvement process. Second, the efficiency of the clustered groups is collectively evaluated, thus allowing for a preliminary efficiency ranking of the different groups. Then, starting with the least efficient group, an iterative evaluation is performed using the improved MDDF model—a process that combines the intracluster assessments and intercluster optimization strategies to derive the iteratively improved paths for the DMUs. Finally, a comparative analysis is conducted of the proposed iterative improvement model, the traditional MDDF model, and the MDDF model with shadow prices. The results indicate that, in terms of input utilization, the proposed model significantly improves DMU efficiency. The application of the proposed model to the efficiency evaluation of 32 Chinese airlines from 2011 to 2020 demonstrates the model′s ability to generate more practical, improved projection paths and provide better performance outcomes.
Yang et al. (Thu,) studied this question.