Abstract: To overcome unclassifiable region(UR) in conventional multi-classification support vector machine,Mahalanobis distance method(MDM) based solution for unclassifiable region is presented.MDM firstly computes the distance between the sample in UR and every dataset,then selects the class with the least Mahalanobis distance for the sample.Experimental results on benchmark datasets show that MDM eliminates UR in experiment data and improves the classification capacity and generalization ability of the algorithm effectively on experimental datasets.