A Hybrid Machine Learning Framework for Enhanced Tsunami Prediction Using Ensemble Models and Neural Networks

Tsunami prediction is a critical task for mitigating risks associated with natural disasters, yet achieving accurate and reliable predictions remains a significant challenge due to the inherent complexity and uncertainty in earthquake-related data. Traditional predictive models often struggle to capture the intricate relationships between earthquake features, such as magnitude, latitude, longitude, depth, and instrumental intensities, leading to suboptimal performance and unreliable predictions. To address these challenges, this research proposes a hybrid machine learning framework that integrates ensemble models and neural networks to enhance both accuracy and robustness in tsunami prediction. The dataset undergoes rigorous preprocessing, including the removal of missing values, normalization, and shuffling, to improve data quality. The framework employs a diverse set of ensemble models such as Random Forest, Gradient Boosting, XGBoost, LightGBM, and CatBoost alongside a neural network with three hidden layers for predictive modeling. Predictions from these models are aggregated into meta-features and passed to a logistic regression meta-classifier for final decision-making. Using ten-fold stratified cross-validation, the framework is evaluated on key metrics, including precision, recall, F1-Score, accuracy, and ROC-AUC. Results demonstrate that the hybrid model significantly outperforms individual models, effectively addressing the challenges of low accuracy and instability in traditional approaches. By leveraging the complementary strengths of ensemble models and neural networks, the proposed framework offers a scalable and adaptable solution for tsunami prediction, contributing to enhanced disaster preparedness and risk mitigation strategies.