Earthquake Precursor Origin and AI Framework for Automatic Classification

Abstract

 Anomalous seismic, environmental, and geophysical signals are examples of observable occurrences that
might serve as earthquake precursors, warning of imminent seismic events. Precise P-wave detection and earthquake
source parameter estimates depend on the accurate identification of these precursors and the separation of
experimental artifacts from real ground-origin signals. At three different sampling rates (50, 100, and 200 samples
per second), a dataset of 429 signals was gathered from the New Abu Dabbab station and classified as either
ramping, non-ramping, or mixed patterns. For the automatic categorization of P-wave signals, a Convolutional
Neural Network (CNN) model was assessed in conjunction with conventional machine learning methods such as
Logistic Regression, K-Nearest Neighbors, Support Vector Machine, Decision Tree, Random Forest, and XGBoost.
With a 97.98% accuracy rate, the CNN model outperformed other models. By incorporating an attention
mechanism, the prediction accuracy increased to 99.81%, indicating improved contextual comprehension and
feature optimization in sequential seismic data. Accuracy, precision, recall, F1-score, and ROC curve analysis were
used in a thorough examination, which validated the deep learning-based method's resilience. According to our
findings, attention-augmented CNN models minimize mistakes in earthquake precursor identification by offering a
dependable and effective framework for real-time P-wave signal categorization.