Prediction of Earthquake Intensity and Location in Sumatra Using Deep Learning

Earthquake prediction is important to help mitigate the risks and damage associated with seismic events, leading to potentially saving lives and reducing economic losses. To predict earthquake events, the Epidemic-Type Aftershock Sequences (ETAS) model is often employed, which analyzes detailed earthquake catalog data by considering a statistical model incorporating seismological laws. However, the ETAS model often fails to predict earthquakes' future occurrences. To overcome this issue, we combine the ETAS model with deep learning methods to enhance prediction performance. We integrate the ETAS model with Long Short-Term Memory (LSTM) and Convolutional Neural Networks (CNN) architectures. By optimizing zone divisions and time-step configurations in the Sumatra region, our implementation demonstrates that this technique significantly improves the accuracy of prediction of earthquake intensity and location, marking substantial progress in seismic research and disaster prevention strategies.