Algoritma Backpropagation untuk Memprediksi Korban Bencana Alam

Abstract

Indonesia is a country prone to natural disasters. Because Indonesia is a maritime country and its geographical area is Mount Merapi. In order to reduce victims of natural disasters or other disasters, we conducted research related to predictions of victims of natural disasters. The purpose of this study is to help the team or related parties in preparing themselves to deal with the victims of a growing natural disaster. The algorithm used in predicting victims of natural disasters is backpropagation. The data used in this study is the DIBI dataset taken from the Google dataset. The predicted impact was 5128 lines, 524 missing victims, 2653 injured, 941 lines dead. Each dataset with each category of disaster impacts, missing victims, injured victims, and death victims was made of 2 input variables. Input variables from each category are district code, and year and the output variable is the number of disaster victims. Neural network structure and architecture of this study, namely 2 input layer nodes, 2 hidden layer nodes, and 1 output layer node. From the architecture, training and testing were carried out, where the results of testing disaster impact data were 110 lines of MSE value of 0.0371, testing results of wounded victims data as much as 53 lines of MSE value of 0.0256, results of testing of missing victims as much as the 24 lines of the MSE value are 0.041, and the results of testing of the dead are 41 lines of the MSE value of 0.029.