Using Support Vector Machine as a Binary Classifier

This paper presents a relatively new and less known alternative of the classical neural network architectures – Support Vector Machines (SVM) and their usage for binary data classification. After a brief description of the Statistical Learning Theory – the framework of SVM, we explore the ways to build an error-tolerant binary classifier for linearly and non-linearly separated data. A comparison between SVM and some other neural network types is performed. To illustrate the paper, a demonstration software is provided.