Perbandingan Klasifikasi SVM dan Decision Tree untuk Pemetaan Mangrove Berbasis Objek Menggunakan Citra Satelit Sentinel-2B di Gili Sulat, Lombok Timur

Abstract

Mangrove is one of the most important objects in wetland ecosystems. Mangrove research has been done, one of them is using remote sensing technology. This study aims to assess accuracy of object based image analysis (OBIA) approach on both Support Vector Machine (SVM) and Decision Tree classification methods to classify mangrove and estimate mangrove area in the field study. We selected Kawasan Konservasi Laut Daerah (KKLD) Gili Sulat as a research site. This research used Sentinel-2B satellite imagery. We took field data using stratified random sampling and the amount of the data we collected were 121 points. The classification analysis result with object based showed that SVM had an overall accuracy of 95 % (kappa = 0.86) and Decision Tree classification had an overall accuracy of 93 % (kappa = 0.82). It is caused SVM can reduce the error of classification than Decision Tree. Estimation result based on assessment showed that mangrove using SVM had 634.62 Ha while using Decision Tree had 590.47 Ha