Sonar Target Classification Problem: Machine Learning Models

Ritwick Ghosh

Abstract: In this study various machine learning algorithms are used for a noisy binary classification problem. The dataset that was used by Gorman and Sejnowski (1988) in their study of the classification of sonar signals using a neural network of undersea targets is used in this study. The task was to train a network to discriminate between sonar signals bounced off a metal cylinder and those bounced off a roughly cylindrical rock. The data used for the network experiments were sonar returns collected from a metal cylinder and a cylindrically shaped rock positioned on a sandy ocean floor. This dataset has 60 different features and is extremely noisy in nature. Total 29 machine classification algorithms are used on the dataset. Programming languages used in this study are MATLAB 2018b and Python 3.7.

Keywords: Sonar, Machine Learning, KNN, Decision Tree, Linear Regression, Ensemble

How to Cite?: Ritwick Ghosh, "Sonar Target Classification Problem: Machine Learning Models", Volume 9 Issue 1, January 2020, International Journal of Science and Research (IJSR), Pages: 247-248, https://www.ijsr.net/getabstract.php?paperid=ART20203916, DOI: https://dx.doi.org/10.21275/ART20203916