Advancing Earthquake Prediction in China: Machine Learning Approaches for Risk Assessment and Magnitude Forecasting

Abstract

Earthquakes pose severe disasters and losses due to their sudden and destructive nature, and despite extensive research, predicting earthquakes remains a significant challenge. Recent advancements in earthquake observation and geophysical methods have highlighted the potential of machine learning technologies in improving prediction accuracy. This study aims to explore earthquake distribution in China through machine learning methods, specifically logistic regression and random forest, to identify high-risk areas and predict earthquake magnitudes. The research identifies five high-risk earthquake zones in China and demonstrates that the random forest model excels in predicting earthquake magnitudes within these zones, outperforming support vector machines and backpropagation models. Notably, the study reveals that the b-value is a crucial factor in earthquake magnitude prediction and should be emphasized in future research. This study not only provides new perspectives and methodologies for earthquake prediction but also offers a scientific basis for earthquake warning and disaster prevention. Future work will focus on incorporating additional seismological and precursor indicators to enhance prediction accuracy and contribute to a deeper understanding and more practical solutions in earthquake forecasting.