PIXEL-BASED IMAGE CLASSIFICATION USING A GREY WOLF OPTIMISED SUPPORT VECTOR MACHINE

Abstract

Support Vector Machine (SVM) is one of the most effective machine learning algorithms that is widely employed for classification tasks. SVMs perform well in high-dimensional spaces, making them suitable for applications with a large number of features. This capability is crucial in tasks like image classification where each pixel can represent a feature. Its effectiveness has made it a preferred choice among remote sensing experts. However, the performance of the SVM is highly dependent on the appropriate selection of the best combination of hyperparameters. Thus, making optimisation an essential step for maximising classification accuracy. This paper explores a metaheuristic optimisation algorithm, the Grey Wolf Optimisation Algorithm (GWO) to optimise the performance of the SVM by fine-tuning the optimal combination of hyperparameters that can improve the accuracy of the SVM. With an accuracy of 92%, the SVM confirms its superiority when optimised with the GWO, as compared to the stand-alone SVM, which obtained an accuracy of 89%. The findings of this research highlight the potential of metaheuristic algorithms in improving the effectiveness of machine learning algorithms for image classification tasks.