Klasifikasi dengan Analisis Diskriminan Fisher, Jarak Mahalanobis, dan Analisis Biplot

Abstract

The classification of a new object into a group is expected to be solved with minimum error. Fisher discriminant analysis, Mahalanobis distance (either with separate or pooled covariance matrix), and biplot analysis are used for classification. The data being used are the Iris and the generated data for simulation. Mahalanobis distance with separate covariance matrix gives the minimum classification error. However, if the classification of a new object into a group is not only based on the minimum classification error but also based on the visualization of the data, Fisher discriminant analysis gives the best result.