Time Series Analysis of Earthquake Activity in Southern Leyte, Philippines: Applications of FB Prophet for Seismic Risk Assessment

Aim/Background This study investigates the challenge of forecasting earthquake magnitudes in Southern Leyte, Philippines, a region vulnerable to seismic activity due to its proximity to major fault lines. Despite advancements in earthquake prediction models, including ARIMA, LSTM, and FB Prophet, accurately forecasting rare, high-magnitude earthquakes remains difficult. This research aims to assess the effectiveness of the FB Prophet time series model in predicting earthquake magnitudes and trends in Southern Leyte. Methods The study utilized 40 years of seismic data (1980 – August 2024) sourced from the USGS earthquake catalog. The methodology involved preprocessing historical data, applying the FB Prophet model, and analyzing forecast components, including trends and periodic variations (weekly, yearly, and daily). Results Results show that FB Prophet successfully captures general seismic trends and periodic fluctuations, with an RMSE of 0.3912, indicating reasonable prediction accuracy for smaller earthquakes. However, the model struggles with forecasting rare, high-magnitude events, as evidenced by wider confidence intervals. Conclusion The study concludes that complementary models, such as LSTM and ARIMA, are necessary to improve prediction accuracy, especially for large seismic events. Future research should focus on integrating hybrid models and additional data sources to enhance forecasting reliability, contributing to better disaster preparedness and resilience in earthquake-prone regions like Southern Leyte.