Assessment of post-disaster building damage levels using back-propagation neural network prediction techniques

Almais, Agung Teguh Wibowo, Fajrin, Rahma Annisa, Naba, Agus, Sarosa, Moechammad, Juhari, Juhari and Susilo, Adi (2024) Assessment of post-disaster building damage levels using back-propagation neural network prediction techniques. JOIV : International Journal on Informatics Visualization, - (-). ISSN 2549-9610 (print) | 2549-9904 (online) (In Press)

Text 20716.pdf - Published Version Available under License Creative Commons Attribution Share Alike. Download (1MB)

Abstract

Indonesia is susceptible to natural disasters, with its geographical location being one of the contributing factors. To mitigate the harmful effects of natural catastrophes, a disaster emergency response must be undertaken, which consists of a set of steps taken immediately following the event. These operations include rescuing and evacuating victims and property, addressing basic needs, providing protection, and restoring buildings and infrastructure. Accurate data is required for effective recovery after a disaster. The Badan Penanggulangan Bencana Daerah (BPBD) oversaw disaster relief efforts, but faulty damage assessments slowed restoration.
Surveyor subjectivity and differing criteria result in discrepancies between reported damage and reality, generating issues during the post-disaster reconstruction phase. The objective of this study is to develop a prediction system to measure the extent of damage caused by natural disasters to buildings. Based on the five criteria that determine the level of building damage after a disaster, namely, building condition, building structure condition, physical condition of severely damaged buildings, building function, and other supporting conditions. The data used are from the BPBD of Malang city from 2019 to 2023. This system would allow surveyors to make speedy and objective evaluations. Five different models were tested using the Neural Network Backpropagation approach. Model A2 produces the highest accuracy of 93.81%. A2 uses a 40-38-36-34 hidden layer pattern, 1000 epochs, and a learning rate of 0.1. These findings can lay the groundwork for advanced prediction models in post-disaster building damage evaluation research.

Item Type:	Journal Article
Keywords:	Predictions; Post-Disaster; Building Damage; Neural Network
Subjects:	08 INFORMATION AND COMPUTING SCIENCES > 0801 Artificial Intelligence and Image Processing 08 INFORMATION AND COMPUTING SCIENCES > 0802 Computation Theory and Mathematics 08 INFORMATION AND COMPUTING SCIENCES > 0806 Information Systems
Divisions:	Faculty of Technology > Department of Informatics Engineering
Depositing User:	Agung teguh Wibowo Almais
Date Deposited:	23 Oct 2024 14:20

Downloads

Origin of downloads

Actions (login required)

View Item