Implementation of Genetic Algorithm (GA) for Hyperparameter Optimization in a Termite Detection System

Abstract

In the development of a termite detection system, four hyperparameters including cost (C), gamma (γ), coefficient (r) and degree (d), must be conscientiously predetermined in establishing an efficient support vector machine (SVM) model. Therefore, the objective of this study is to develop a robust classification model generated by a genetic-based SVM (GA-SVM) that can automatically determine the optimal parameters of SVM with the highest predictive accuracy and generalization ability. Based on acoustic signals, the energy and entropy are derived as a feature input to the SVM classifier to detect termites. From these experimental results, it can be seen that the GA-SVM can more significantly improve the performance of our proposed system compared to previous research based on the grid-search method. Based on the numerical analysis, our proposed system achieves the excellent accuracy of 0.9264.