Human Development of Countries Using Mixtures Discriminant Analysis

Linear Discrimination Analysis (LDA) is based on normality assumption. When this condition is not the case, nonparametric methods have to be used. In this paper it is assumed that the classes are not normal, but each class includes subclasses with the normal distribution. This leads to mixture discrimination analysis, MDA, and EM algorithm needed for classification between classes. Then, the Human Development data in 2002 has been used and MDA software is adapted to classified countries into three categories. It has been shown that the proposed method has more performance than LDA method.