Support Vector Machines

This chapter describes support vector machines (SVMs) and illustrates their use in microbiome data using the e1071 and microbiomeMarker packages. In addition to the Summary, there are five sections in this chapter. Section 8.1 introduces separating hyperplanes, including definitions of a hyperplane and maximum margin classifier, which is the basis for support vector classifiers. Section 8.2 introduces and defines support vector classifiers and their construction. Section 8.3 describes SVMs via two techniques: dot product and kernel function (kernel trick), and introduces how to construct SVMs for classification and kernel methods for Gaussian and binary outcomes. Section 8.4 implements SVMs with the e1071package to illustrate the algorithms and with the microbiomeMarker package to identify the important microbiome markers. Section 8.5 provides some remarks on SVMs, including its algorithms, advantages, and disadvantages as well as model complexity and parameter setting.