The Visualization and Classification Method of Support Vector Machine in Lymphoma Cancer

Abstract

In the classical-classification multivariate process, it becomes an interesting topic to be discussed in the research area because of the larger variables with smaller observations. For this we need a method that can handle this problem. One answer is to use machine learning. SVM is a classification method in machine learning that is able to classify these data types. In addition, SVM can also model and classify relationships between variables efficiently and easy interpretation. This paper aims to create a visualization of SVM classifiers, then obtain an accuracy value to have an optimal classification with a misclassification of small numbers. This study aims to find good SVM input parameters by assessing the importance of variables using visual methods. This visualization will distinguish groups of people who contract diffuse lymphoma cancer and follicular lymphoma cancer with data on the genetic expression of lymphoma cancer. The classification using kernel Linear, Gaussian RBF, Polynomial and Sigmoid. The best classification accuracy using linear kernel functions with training data has a classification accuracy of 100% and testing data has a classification accuracy of 94, 73%.