Architectural Models for Predicting the Amount of Natural Disasters and their Effects Using Batch Training

N. L.S.R. Ginantra, S. Kesuma, G. S. Achmad Daengs, G. W. Bhawika, N. Mulyani, I. Hasian, Y. Siagian

Research output: Contribution to journalConference articlepeer-review

Abstract

The Batch Training method is one of the methods of Artificial Neural Networks that can be used to make predictions, especially in times series data. This method is able to make predictions by learning from data that has never happened before by forming the right network architecture model. Therefore, this research will discuss the best network architecture model that is appropriate for making predictions using the Batch Training method. The data used in this study is the data of Natural Disasters in Indonesia, sourced from the National Disaster Management Agency. There are 12 variables used, namely Time of disaster, Number of disasters, Death and missing victims, injured victims, victims suffering and displaced, seriously damaged houses, lightly damaged homes, submerged houses, damage to health facilities, damage to worship facilities, and damage to facilities education. Based on this data will be formed and determined the network architecture model used, including 4-5-1, 4-10-1 and 4-15-1. From these 3 models after training and testing, the best architectural model is obtained 4-10-1 with an accuracy level of 91% with MSE Training and testing values of 0.0245532940 and 0.0579265906.

Original languageEnglish
Article number012032
JournalJournal of Physics: Conference Series
Volume1566
Issue number1
DOIs
Publication statusPublished - 3 Jul 2020
Event4th International Conference on Computing and Applied Informatics 2019, ICCAI 2019 - Medan, Indonesia
Duration: 26 Nov 201927 Nov 2019

Fingerprint

Dive into the research topics of 'Architectural Models for Predicting the Amount of Natural Disasters and their Effects Using Batch Training'. Together they form a unique fingerprint.

Cite this