Machine Learning-Based Prediction Models to Improve the Accuracy of Early Earthquake Detection in Cities: A Systematic Review (#863)

Authors

Espiritu Bustamante, Anthony Gerard

Carhuapuma Amanqui, Catherin Karla

Ayala Ñiquen, Evelyn Elizabeth

Yactayo Arias, Cesar Augusto

Abstract

Earthquakes cause significant losses, which demands more efficient strategies for early detection and damage assessment. Given the limitations of traditional methods, this Systematic Literature Review (SLR) aimed to analyze Machine Learning (ML) models applied to seismology to strengthen urban seismic risk management. A rigorous search was conducted in Scopus and Web of Science, yielding 335 articles. After applying inclusion/exclusion criteria and filters, 32 final articles were selected. The results revealed that algorithms such as Convolutional Neural Networks (CNN), Support Vector Machines (SVM), Random Forest, Long Short-Term Memory networks (LSTM), and Artificial Neural Networks (ANN) show great potential in improving the accuracy of early detection of seismic events (P-waves, hypocentral parameters) and in the estimation of structural damage, thereby optimizing response efficiency. However, challenges were identified regarding data availability and quality, as well as model generalization. In conclusion, ML models are a promising tool for urban seismic management, and it is crucial to address existing barriers and explore future research directions to maximize their impact.