Effective Land Damage Classification using Deep Neural Network Method

Abstract

Tsunami disasters can cause substantial damage to land and residential areas, making rapid and accurate damage classification important for disaster assessment. This study applied a Deep Neural Network (DNN) method to classify tsunami-related land damage into three categories: Severe, Light, and Moderate. The dataset consisted of disaster impact records from 34 provinces in Indonesia during 2013–2018. The input variables included the number of deaths, injured people, affected people, and houses with severe, moderate, light, and submerged conditions. We divided the data into training and testing datasets and evaluated the model using accuracy, precision, recall, and F1-score. The proposed DNN achieved an overall accuracy of 83.33% on 18 testing samples. The Moderate category produced the best performance, with precision, recall, and F1-score of 1.00. The Severe category achieved a recall of 1.00 and an F1-score of 0.73, while the Light category obtained a precision of 1.00, recall of 0.62, and F1-score of 0.77. The confusion matrix showed that the model correctly classified all Severe and Moderate samples but misclassified three Light samples as Severe. These findings demonstrate that the DNN can effectively classify tsunami-related land damage, although additional and more diverse data are required to improve the distinction between Light and Severe damage categories.