Comparison of Support Vector Machine and Random Forest Methods on Sentinel-2A Imagery for Land Cover Identification in Banda Aceh City Using Google Earth Engine

Abstract

Land cover is a physical feature of the earth that illustrates the relationship between natural processes and social processes. Over time, there has been a lot of land conversion, where initially open land is now built-up land. This is due to the large-scale development in Banda Aceh City. Therefore, this study aims to compare the performance of two classification methods, namely using Support Vector Machine (SVM) and Random Forest in identifying land cover in Banda Aceh City using Sentinel-2A imagery via the Google Earth Engine platform. As for data recording, it starts from January 1 to December 31, 2023. There are 4 classes used in this study, namely vegetation, water bodies, built-up land, and open land. The classification results show that the Support Vector Machine and Random Forest methods have been successfully applied to identifying land cover in Banda Aceh City using Sentinel-2A imagery. The accuracy results show that the Support Vector Machine method has a higher accuracy value of 90.5% compared to the Random Forest method of 85.7%.