Data envelopment analysis clustering approach with influence domains
针对数据包络分析聚类结果不稳定、无法有效划分分段前沿交点及边界僵化的问题，引入影响域概念并叠加影响域来调整聚类，提出聚类约简和质量测量方法，经案例验证有效。
As a novel means of clustering, data envelopment analysis (DEA) clustering approach groups decision-making units (DMUs) based on similar production functions. However, the method faces unstable clustering results and cannot effectively categorise the intersection points of piecewise frontiers. More importantly, the existing methods are limited by rigid cluster boundaries, which fail to reflect the variations in the frontier. Therefore, the paper introduces the concept of influence domains into the general process of the DEA clustering approach for solving these problems. Meanwhile, by using the overlay influence domain, clusters can be reasonably adjusted to better reflect the frontier’s variations. Then, order differences are assigned to the influence domain to further distinguish the preferences of DMUs within the domain. Subsequently, a clustering reduction method is proposed to optimise the number of clusters, along with a DEA clustering quality measurement method. Finally, case studies confirm the effectiveness of the proposed methods.