Klasifikasi Wilayah Potensi Risiko Kerusakan Lahan Akibat Bencana Tsunami Menggunakan Machine Learning

Abstract

Indonesia is an archipelagic country with a long coastline where some areas are prone to tsunami waves which can result in land damage. Tsunamis occur due to earthquakes or volcanic eruptions under the sea that cause movement of the seabed and then create strong waves. The Special Region of Yogyakarta, precisely in Bantul Regency, is one of the areas that have a high risk of a tsunami disaster because the area is located in the expanse of the Indian Ocean which has quite impulsive plate movements. This study aims to find out information about the level of risk of land damage due to the tsunami using vegetation index data from OLI 8 Landsat imagery. Classification or prediction using the Artificial Neural Network (ANN) method. The vegetation index used is NDVI, NDWI, NDBI, SAVI, and MNDWI. The relationship between SAVI and NDVI has a positive correlation coefficient with the highest value of 0.962 where the potential risk of low damage is 0.933 and the potential risk of high damage is 0.856. Classification of potential areas of high risk of damage to tsunami land (High Risk) using the ANN method resulted in 7 villages with high risk. The ANN algorithm is the most accurate method for classification predictions between the Random Forest and SVM methods which get an accuracy of 95.45% and get a Kappa value of 86.08%. Spatial prediction using IDW produces a map of the distribution of the potential risk area for land damage caused by the tsunami.