Smart Seismic Intelligence Machine Learning for Spatial Clustering and Earthquake Magnitude Prediction in Indonesia

Abstract

Indonesia is located within the Pacific Ring of Fire, one of the most seismically active regions in the world due to the interaction of multiple major tectonic plates. Understanding the spatial distribution of earthquakes and accurately estimating their magnitudes is essential for effective disaster risk assessment and mitigation planning. This study aims to analyze earthquake distribution patterns and develop a machine learning-based approach to predict earthquake magnitude using seismic data from the Meteorology, Climatology, and Geophysics Agency (BMKG). The study employs two machine learning methods: K-Means Clustering to identify spatial groupings of earthquake events and Random Forest Regression to predict magnitude based on spatial and temporal features. The dataset consists of 67 earthquake events recorded in February 2026, including attributes such as latitude, longitude, depth, magnitude, and occurrence time. Clustering results indicate that the optimal number of clusters is k = 4, with a Silhouette Score of 0.3444, suggesting a moderate clustering structure. This implies that spatial patterns are present, although cluster separation is not yet well-defined. The Random Forest model achieved an R² of 0.7382 on training data and 0.0975 on testing data, indicating overfitting likely due to the limited dataset size. Feature importance analysis reveals that longitude contributes the most (43.7%), followed by depth (29.6%), latitude (20.6%), and time (6.0%). These findings highlight the dominant role of spatial factors in Indonesia’s seismic activity. However, the limited dataset restricts model generalization; therefore, future studies should use larger datasets and incorporate additional geophysical parameters to improve predictive performance.