Evolving Machine Learning-Based Classifiers by Metaheuristic Approach for Underwater Sonar Target Detection and Recognition

Abstract

Nowadays, considering their acoustic radiated signal, the classification of underwater sonar targets encompasses a broad spectrum of targets and techniques. However, classifying various sonar targets, such as maritime vessels, is difficult for academics, who must consider both military and commercial considerations. Support Vector Machines (SVMs) are the most operational type of machine learning model for creating classifiers. The most difficult aspect of SVM networks is the learning algorithms.Therefore, in this chapter, in order to have a reliable and accurate underwater sonar target classifier, we investigate the performance of ten metaheuristic algorithms, including slime mould algorithm (SMA), marine predator algorithm (MPA), Kalman filter (KF), Harris Hawks optimization (HHO), genetic algorithm (GA), particle swarm optimization (PSO), Henry gas solubility optimization (HGSO), chimp optimization algorithm (ChOA), gray wolf optimizer (GWO), and whale optimization algorithm (WOA), in underwater sonar target classification problem.