AI for Earthquake Prediction: A Comparative Analysis of Machine Learning Techniques in Natural Disaster

Abstract

Abstract—Earthquakes are complex natural disasters that may significantly impact people's well-being, their possessions, and the natural world. Inaccurate estimates of earthquake time, location, and magnitude are common since earthquakes do not follow any particular patterns. The capacity of AI-based techniques to uncover previously unseen patterns in data has made them famous. This research presents an AI-driven method for earthquake prediction using the Gated Recurrent Unit (GRU) model and historical seismic data from the USGS. After the data is prepared, features are chosen, and the model is trained, the accuracy, precision, recall, and F1-score are used to assess the performance. The GRU model outperforms Logistic Regression (LR), Recurrent Neural Networks (RNN), and Artificial Neural Networks (ANN) in a comparative comparison. It successfully achieves an accuracy of 93.10% while minimizing overfitting. The results highlight the effectiveness of GRU in capturing temporal dependencies in seismic events. However, challenges such as data imbalance, computational complexity, and regional generalization remain. Future research should focus on integrating additional geological and environmental parameters, optimizing computational efficiency, and developing real-time predictive frameworks to enhance the reliability of AI-driven earthquake forecasting.