Hierarchical Discriminant Analysis and Its Application

This article presents a new discriminant analysis method and tries to improve the accuracy of discriminant problems without reducing variables. This method establishes two layers of discriminant functions. For the first layer, a linear discriminant function is constructed to determine the sample type. Those samples’ type are not determined at the first layer, they will be distinguished as sub-samples, then a second layer of a nonlinear discriminant function will be established to determine the sub-sample type. The examples show that this method can handle discriminant problems of complex data, and results of this method are better than some common methods.