Teknik Decision Tree dalam Pengklasifikasian Penggunaan Lahan dengan Menggunakan Citra Sentinel-2A MSI

Abstract

Land use classification is an interesting thing to study and research. There are several classification methods that are widely used in some researches, one of which is using machine learning methods, such as SVM, Naive, and decision trees for both aerial photography and satellite imagery. This study uses MSI Sentinel-2A satellite imagery to classify the land use of Kota Langsa as the study area. As for the method used, the first step is downloading the Sentinel-2A image with a band that has a resolution of 10 m and 20 m. This image has an extremely good resolution compared to other optical remote sensing images. For the second step, because the image of the earth's surface is covered by clouds, it is necessary to do a cloud masking process to reduce pixel classification errors. Next, the process of collecting training datasets obtained from each class and represented by pixel values. Then, the training dataset is used to perform a supervised classification using the decision tree method. This method will categorize each pixel into eight classes. As a validation stage, it is necessary to obtain the accuracy of the classification results by using a confusion matrix and calculating the overall accuracy (OA). From the accuracy, it can be concluded that the decision tree can give good results in land use classification of Kota Langsa with 94% of OA value.