Klasifikasi Zona Tanam Mangrove Desa Pulau Sembilan Menggunakan Gaussian Process Regression for Classification (GPRC)

Abstract

Over the past two decades, mangrove forests have undergone substantial deforestation. North Sumatra Province, as one of Indonesia's largest repository of mangrove forests, covering approximately 57,490 hectares, primarily situated along the East Coast of Sumatra. Notably, Pulau Sembilan Village, nestled within Pangkalan Susu District, Langkat Regency, North Sumatra, emerges as a notable hub of mangrove growth, featuring a thickness range spanning from 100m to 1700m. Despite its considerable expanse, this ecosystem faces imminent threats posed by human-induced activities such as illicit logging, waste accumulation, and the overarching impacts of global climate change. Consequently, this study employs the Gaussian Process Regression for Classification (GPRC) methodology to delineate mangrove planting zones within Pulau Sembilan Village, Langkat Regency, North Sumatra. Leveraging water content data alongside diverse environmental features, the GPRC model undergoes training employing three distinct kernels: Radial Basis Function (RBF), Matern, and Rational Quadratic. The training regimen entails 25 iterations per kernel, with performance assessments conducted via accuracy, F1-Score, precision, and recall metrics. Outcomes indicate the Rational Quadratic kernel yielding the highest accuracy at 98.29%, followed by the Matern kernel at 98%, and the RBF kernel at 96.58%. Subsequent testing underscores the superiority of the Rational Quadratic kernel, attaining an accuracy rate of 98.15%. These findings underscore the efficacy of GPRC in classifying mangrove planting zones, with the Rational Quadratic kernel emerging as particularly adept. Ultimately, this research contributes valuable insights towards mangrove planting zone management and offers a robust foundation for future investigations into GPRC-based classification methodologies.