Implementation of Logistic Regression Algorithm in Predicting Tsunami Potential on Earthquake Data Parameters

This study presents the evaluation and testing of a logistic regression model for predicting earthquake-related features, including earthquake depth, magnitude, and tsunami potential. The model achieved high accuracy in predicting earthquake depth categories (99.82%) and earthquake magnitude (99.84%), but faced challenges with low recall for tsunami prediction (50%) due to class imbalance. Evaluation results showed that the model struggled to predict tsunami occurrence accurately, as the dataset contained a disproportionate number of 'no tsunami' instances. Despite these limitations, the model displayed high accuracy for earthquake depth and magnitude predictions. The testing phase revealed a series of prediction errors, particularly for the tsunami category, influenced by the imbalance in training data. The results emphasize the need for improved handling of imbalanced datasets and the potential for exploring other machine learning algorithms and techniques for better performance in multiclass classification problems. Future research could further refine these models by incorporating additional criteria and exploring other earthquake and tsunami prediction methodologies.