DEA에서 교차효율성의 공격적 정형화

We propose a new aggressive formulation of cross-efficiency in Data Envelopment Analysis (DEA). In the traditional aggressive formulation, the efficiency score of a test DMU is maximized as the first goal while an average of efficiency scores of peer DMUs is minimized as the second goal. The proposed model replaces the second goal with the minimization of the best efficiency score of peer DMUs. We showed the model is a quasi-convex optimization problem, and for a solution method we developed a bisection method whose computational complexity is polynomial-time. We tested the model on 200 randomly generated DEA problems, and compared it with the traditional model in terms of various criteria. The experimental results confirmed the effectiveness and usefulness of the proposed model.