The standard data envelopment analysis (DEA) method assumes that the values for inputs and outputs are exact. While DEA assumes exact data, the existing imprecise DEA (IDEA) assumes that the values for some inputs and outputs are only known to lie within bounded intervals, and other data are known only up to an order. In many real applications of DEA, there are cases which some of the input and output variables belong to bounded and discrete sets and the others are known to lie within bounded intervals. In this paper a new variety of imprecise data in DEA has been faced. The proposed approach transforms a nonlinear DEA model to an equivalent linear programming problem. Upper and lower bounds of efficiency scores of operational units are defined. A real case on commercial banks is provided to illustrate the applicability of the approach.