<![CDATA[Monitoring and classification of forest land cover on small islands require accurate and efficient methods to support sustainable natural resource management. This study aims to evaluate the effectiveness of Machine Learning algorithms, namely Classification and Regression Tree (CART), Support Vector Machine (SVM), and Random Forest (RF), in classifying forest land cover on Nusalaut Island, Maluku Province, and to compare their performance in terms of accuracy, efficiency, and computational resource requirements. The study utilized Sentinel-2 Level-2A satellite imagery from 2025, processed using the Google Earth Engine platform. Supervised classification was applied to four land cover classes, namely water bodies, built-up areas, open land, and vegetation. Model performance was evaluated using a confusion matrix to obtain Overall Accuracy (OA) and the Kappa coefficient. The results indicate that all three algorithms produced high classification accuracy, with SVM and RF achieving the best performance, attaining an OA of 98% and a Kappa value of 0.97, while CART achieved an OA of 94% and a Kappa value of 0.90. SVM demonstrated superior class separation for land cover types with distinct spectral characteristics, whereas RF was more robust to data noise. These findings suggest that Machine Learning algorithms, particularly SVM and RF, are highly effective for forest land cover classification in small island environments such as Nusalaut Island.]]>
Copyrights © 2026
