ARTIFICIAL NEURAL NETWORK BASED EARTHQUAKE INJURIES ESTIMATION MODEL MODEL

• Main Author: Imaduddin Azhar, Ihsan
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/62136
• Institution: Institut Teknologi Bandung

In this study, an artificial neural network model was developed to estimate the number of earthquake victims. The model development was carried out with three datasets, namely the earthquake occurrence dataset in Istanbul, China, and a combination of Southeast Asia and South Asia. The search for ANN architecture was carried out by particle swarm optimization with MAE from the model as an objective function. The experimental results show that the model developed with the earthquake incident dataset in Istanbul has an MAE of 338.4 with an R2 of 0.53, the model developed with the earthquake event dataset in China has an MAE of 63.3 with an r2 of 0.36, and the model developed with the earthquake event dataset in Southeast Asia and South Asia has MAE 4513.6 with R2 -0.003. From the results of the analysis, the lack of features and the small size of the dataset are the reasons for the poor performance of the model. While the results of the search for model architecture with PSO show that PSO can speed up the process of searching for an artificial neural network model architecture by compensating the model's performance when compared to the exhaustive method.