Comparative Performance of Classical Fisher Linear Discriminant Analysis and RobustFisher Linear Discriminant Analysis

Linear discriminant analysis for multiple groups can be performed usingFisher s technique which can be applied to classify and predict observations into variouspopulations. Classical Fisher linear discriminant analysis (FLDA) is highly susceptible tooutliers. The poor performance of classical FLDA is due to lack of robustness of theclassical estimators used to train the model. The proposed robust FLDA combine thefeatures of classical FLDA and weighted sample observations. This paper examines thecomparative classification performance of Fisher linear discriminant analysis and theproposed robust Fisher linear discriminant analysis. The paper focuses on the influencescaled normal and unscaled normal data set have on the classical Fisher and the robustFisher techniques. The objectives of this paper are to compare the classificationperformance of these methods based on the mean of correct classification and to examinethe separation between the group means. The classification results indicate that theproposed procedure has improved classification rate compared to the classical Fisherlinear classification analysis. The simulation showed that both procedures havecomparable separation capability.