Geological Time Series Analysis for Spatial Seismic Forecasting in Japan

Earthquakes pose significant threats to lives and infrastructure worldwide, underscoring the critical need for effective prediction models. Our study represents a meticulous examination of seismic data sourced from the United States Geological Survey, encompassing earthquake occurrences in Japan over five years. Employing a refined zoning approach based on historical seismicity, we tailored our analysis to account for regional variations in earthquake frequency and intensity. By integrating these discoveries into the ARIMA model, we aim to fortify earthquake prediction frameworks, thereby fostering more effective disaster mitigation strategies on a global scale. Furthermore, our analysis using SMA, EMA, and ARIMA models demonstrated that the EMA model provided superior performance. Leveraging the predictive capabilities of the EMA model, we identified twelve geographic coordinates in Japan with a higher likelihood of future earthquakes. This study underscores the potential of EMA in earthquake prediction and provides a foundational approach for integrating machine learning with geospatial analysis for natural disaster forecasting.