Comparing the Performance of the Kernel Functions in the LDA-SVM Based Classification Algorithm in the LabVIEW Environment

Abstract

Machine learning, which is a paradigm of an unknown method that makes inferences of existing data by mathematical and statistical methods, consists of different algorithm choices according to the structure of the used data. Machine learning consists of supervised-unsupervised learning algorithms and dimension reduction algorithms with advantages and disadvantages within themselves. Linear Discriminant Analysis (LDA) is the most commonly used as dimensionality reduction technique for pattern recognition and machine learning applications. Also, Support Vector Machine (SVM) is one of the most used supervised algorithms because of its ability to separate data from different classes from each other by a remote hyperplane, high performance, easy to implement and high generalization capability. In this paper, the dimension of breast cancer data was reduced by LDA and the obtained feature data was classified by using SVM. In order to compare the performance of LabVIEW with MATLAB, data reduced in dimension in LabVIEW and feature data is classified with codes written in the MATLAB and LabVIEW. Then, model performance results are analyzed with LabVIEW. In SVM, Gaussian and Polynomial kernel functions are used and obtained results are compared with each other.