There are several variable selection methods for deciding which variables to include in discriminant analysis. The purpose of variable selection techniques is to choose a suitable subset of variables. There are three common methods are usually referred to as forward selection, backward elimination and stepwise method. The linear discriminant analysis has long been known since fisher, 1936 and can be used not only to examine multivariate differences between groups, but also to determine which variables are the most useful for   discriminant between groups. In this paper, a new approach will be introduced to select the most important variables in discriminant function using mathematical programming (MP). The mathematical programming approach used to discriminant between two or more than two groups. The new selection approach can be applied directly to discriminant function with respect to their parameters. which helped to select the desirable number of variable like the variable selection techniques. This variable used to be binary variable equal one when the variable was selected otherwise equal zero when another variable wasn't selected. The suggested mathematical programming approach doesn't have the assumptions of the multivariate statistical techniques. The suggested approach has been used successfully in a number of applications that are described