The technique of data envelopment analysis (DEA) introduced by Charnes, Cooper and Rhodes (CCR) in 1978 has been widely applied to evaluating the relative efficiency of decision making units (DMUs). DEA provides not only efficient performance of each assessed DMU but also a target that improves efficiency of the DMU. The efficient targets provided by the classical DEA models are always very far from the assessed DMU. However, the closest efficient target is often more appropriate because it needs less effort to make the DMU efficient from the perspective of managers of DMUs. The difficulties of computing the closest efficient target are: (a) the definition of the efficient frontier is given in an implicit fashion, that is hard to be exploited in an algorithm; (b) the efficient frontier is nonconvex. In our research, in order to overcome these difficulties, we use the optimization tool (Karush-Kuhn-Tucker conditions) to transform the definition of the efficient frontier and make the definition computation-friendly. The main works we have done can be summarized as follows.

(1) We proposed a new approach that can provide an efficient target that is closer to the assessed DMU than that provided by the existing studies;

(2) We used the proposed approach in (1) to assess the bankruptcy-based performance of Japanese banks. Then, an early warning of the firm's financial performance and an easy-to-achieve improvement plan for the default firm can be provided.