Predictive Modeling of Earthquake Magnitude and Occurrence in Thailand Using Time Series and Geospatial Data

This study explores the temporal and spatial dynamics of earthquake activity in Thailand using observational data recorded between 2020 and 2025. We analyzed over 100 seismic events to investigate magnitude patterns, recurrence intervals, and regional clustering. Using a combination of statistical and deep learning models—including ARIMA and Long Short-Term Memory (LSTM) networks—we developed predictive frameworks to estimate the magnitude of future earthquakes. The LSTM model achieved a mean absolute error (MAE) of 0.61, outperforming ARIMA and Random Forest benchmarks. Kernel Density Estimation (KDE) revealed persistent seismic hotspots in the northern and western provinces, particularly near the Mae Chan and Three Pagodas Fault Zones. These findings suggest that even in data-limited environments, meaningful short-term forecasting and risk zoning can be achieved to support Thailand’s regional early warning systems and disaster mitigation planning.