Machine Learning Framework for Automatic Classification of Nano-Seismic Signals Preceding Earthquakes

Authors

Ghada Ali, Seismology Department, National Research Institute of Astronomy and Geophysics (NRIAG), Helwan 11421, Egypt AND Electrical Engineering Department, Faculty of Engineering, Aswan University, Aswan, EgyptFollow
Ali G. Hafez, Seismology Department, National Research Institute of Astronomy and Geophysics (NRIAG), Helwan 11421, Egypt; Control and Computer Department, College of Engineering, Almaaqal University, Basrah, Iraq AND R&D Division, LTLab, Inc., 1-30-3 Higashi-aburayama, Jonan-ku, Fukuoka 814-0155, JapanFollow
Sayed Hasaneen, Electrical Engineering Department, Faculty of Engineering, Aswan University, Aswan, EgyptFollow
Hamed Nofel, Seismology Department, National Research Institute of Astronomy and Geophysics (NRIAG), Helwan 11421, EgyptFollow
Nabeel Al-Atwan, Control and Computer Department, College of Engineering, Almaaqal University, Basrah, IraqFollow
Ahmed Mohamed, Electrical Engineering Department, Faculty of Engineering, Aswan University, Aswan, EgyptFollow

Abstract

Nano-seismicity preceding earthquakes has many applications in fault prediction, in structure health monitoring, and in enhancing fault and source parameter determination. Not all earthquakes are preceded by these precursors; therefore, there is a need to identify P-wave arrivals that are preceded by these nano signals, which will eliminate errors in P-wave arrival timing. The current work introduces topologies capable of automatically categorizing the arrivals with nano-seismic precursors. The proposed methods are trained on datasets extracted from stations belonging to Egyptian National Seismic Network (ENSN). These records contain both P-wave arrivals preceded by such nano precursors and other records without these signals. The training methodologies are based on several machine learning (ML) models to highlight the elements that distinguish each pattern. The proposed classifier topologies recognize these patterns, and the automated P-wave detector can decide when the arrival of the P-wave exists depending on the detection of the precursors. The examined classification topologies are Logistic Regression (LR), K-Nearest Neighbors classifier (KNN), Support Vector Machine (SVM), Decision Tree Classifier (DT), Random Forest Classifier (RF), XGB Classifier, Naïve Bayes (NB), and Voting Classifier. Out of these methodologies, the SVM outperformed with a classification accuracy of 89.67% because of its capability in feature extraction.

Recommended Citation

Ali, Ghada; Hafez, Ali G.; Hasaneen, Sayed; Nofel, Hamed; Al-Atwan, Nabeel; and Mohamed, Ahmed (2025) "Machine Learning Framework for Automatic Classification of Nano-Seismic Signals Preceding Earthquakes," Almaaqal Journal of Sustainability and Emerging Technology: Vol. 1: Iss. 1, Article 3.
Available at: https://ajset.almaaqal.edu.iq/journal/vol1/iss1/3