Multinomial Classification of Coral Species using Enhanced Supervised Learning Algorithm

Abstract

A supervised learning algorithm can be categorized into different forms and one of this is classification where the main goal is to predict the categorical class labels of structured or unstructured data. However, it requires large datasets to produce a good computer vision model. This study demonstrates the application of the supervised learning algorithm named Convolutional Neural Network (CNN) in multinomial classification of coral reef species. Through the backpropagation process of CNN, the model is able to learn the weights that yield accurate outputs. Moreover, data augmentation approach, retraining, fine tuning and optimization are used to provide better results in multi-class classification. The classification result in terms of F1 Score and Sensitivity is equal to 1.0 while validation accuracy yields 99.49 percent after nine (9) epochs applied to the various coral reef species available in the dataset used in this study.