Mixed integer programming approach of extended data envelopment analysis–discriminant analysis

Several unique features of data envelopment analysis (DEA) are incorporated into discriminant analysis (DA). Such a combined use of the two approaches is referred to as “DEA–DA”. Sueyoshi has recently extended DEA–DA in a manner that it can deal with a negative value in data. The new version is referred to as “Extended DEA–DA”. An important feature of extended DEA–DA is that it can estimate weights of a linear discriminant function by minimizing the total deviation of misclassified observations. As an extension of his research, this research proposes a new type of mixed integer programming formulations that provide these weight estimates by minimizing the total number of misclassified observations.