Comparison Robust Kennel Discriminant Analysis with Kennel Discriminant Analysis by Simulation

Abstract

The majority of data in the real world deviates from the ideal assumptions made by standard statistical methods. In these situations, the assumption of normality in the data (normal distribution of the data) is violated, or there are non-linear aggregate structures in the data set. We are facing a problem in classification. Traditional discriminant analysis cannot confront this problem. We must search for a robust method that deals with this problem. Therefore, this thesis aimed to use the Robust Kennel Discriminant analysis (RKDA) method in the case of contamination in the data and compare it with discriminant analysis. Using the classification error rate criterion to choose the best band width using an accuracy criterion, as it was shown that the robust kennel discriminant analysis method has an advantage over other methods when density functions deviate from the normal distribution with a high percentage of preference.