Support vector machines a evoluční algoritmy

Abstract

The purpose of this work is to create an alternative calculation Support Vector Machines (SVM) using evolutionary algorithms. Normally, the position of the hyperplane at SVM is calculated using the distance from the support vectors. In this work, the hyperplane will be placed stochastically and, using evolutionary algorithms, will evolve to be placed in the most appropriate position with the most appropriate direction. The work contains a comparison of known types of evolutionary and swarm algorithms such as genetic algorithm, differential evolution, particle swarm optimization, gray wolf optimization and its improved version. The work compares algorithms on four different datasets with different structure to show which problems it is appropriate to use the algorithm. This means that not only the datasets are thoroughly analyzed, but also the results achieved using both the standard SVM algorithm and various evolutionary algorithms.