LSTM-Based Forecasting of Seismic Events Using Geomagnetic Indices with Balanced Training

Abstract

Earthquake forecasting remains a notorious challenge due to the complex and seemingly unpredictable nature of seismic events. Recent studies have explored potential precursors in geophysical signals such as geomagnetic indices, with mixed evidence on whether solar–terrestrial interactions influence earthquakes [1,2]. In this paper, we propose a Long Short-Term Memory (LSTM) model to predict the occurrence of a significant earthquake within the next seven days from the global geomagnetic Dst and Kp indices. The model ingests 28-day sequences (224 time steps) with eight derived features and outputs a daily binary alert. To address class imbalance, we combine downsampling and class weighting and tune the decision threshold. Evaluated prospectively on a held-out 2023 period under the natural event prevalence, the model attains F1 = 0.94, ROC-AUC = 0.96, and Accuracy = 0.81, with an explicit precision–recall trade-of. Our findings are exploratory: we do not claim causal links, and results are reported for a single three-year global dataset. Nevertheless, the signal appears modest and potentially complementary to seismic approaches. We discuss threshold policies, possible integration into early-warning workflows, and limitations regarding physical interpretation and generalization.

More by Marat Nurtas

• Two-stage synthetic-to-real transfer learning for automated mammography report generation using vision-language modelsScientific Reports · 2026
• Macro-scale earthquake forecasting under class imbalance in Central AsiaJournal Of Big Data · 2026
• Enhanced Fine-Tuning Optimization for Real-Time Fire and Smoke Detection Using a YOLOv9 Architecture with a Comprehensive Ablation AnalysisEngineering Technology & Applied Science Research · 2026 · Citations: 1
• Spatially honest validation of machine-learning prospectivity maps for karst bauxite in KazakhstanResearch Square · 2026