Classification performance of machine learning methods in different data structures

Abstract

With the popularity of machine learning methods in many areas, the use of computers for diagnosis and treatment in the health field has recently become more frequent. Dual classification studies for the diagnosis of the presence of the disease are quite common in the literature. In this study, classification performances of machine learning algorithms were compared in cases where the response variable is in ordinal structure and more than two categories, instead of binary classification. In the simulation study, data sets in different structures were derived and classification was made. The response variable in the study is an ordinal categorical variable. A comprehensive classification study was carried out using five different machine learning methods. The results show that the SVM method performs better classification than its competitors when the response variable is ordinal.