Abstract: The traditional Data Envelopment.A nalysis(DEA) models are unable to discriminate between ef ficient DM Us. This issue has become the interest of many DEA researchers. AP model is a popular super-efficient model for ranking efficient DMUs, but it has infeasibility and instability problems. MAJ model and LJK model recently proposed can overcome these deficiencies but their ability to distinguish efficient DMUs is not strong. This paper analyzes the reason of these twomodels' shortcomings, then improves them and provides theoretical proofs.Through two standard numerical examples, we find the improved models perform better than old ones, especially the improved LJK model.

Key words: supper-efficient model, MAJ model, LJK model, improvement