Data envelopment analysis (DEA) has been proven as an excellent data-oriented efficiency analysis method for comparing decision making units (DMUs) with multiple inputs and multiple outputs. In conventional DEA models, it is assumed that the input or output variables are all non-negative and desirable. However, in some situations, a performance measure can take positive quantity for some DMUs and negative value for others. Also, undesirable (bad) inputs and outputs may be presented in the production process. Hence, the standard model cannot directly reflect the efficiency score. The paper proposes a modified model in which both undesirable and negative data are treated to improve the relative efficiency of the DMU under evaluation. The focus of this paper is on treating the negative data on the definition of the two non-negative variable and the decreasing of undesirable outputs. A real example of 20 bank branches shows applicability of the proposed approach Manuscript profile