Machine Learning-Based Ionospheric Anomaly Detection for Earthquake Prediction in Northeast Region, India

This study presents a novel machine learning-based framework for earthquake forecasting in the North-Eastern Region (NER) of India, with a focus on ionospheric data collected from Manipur—a region known for its high seismic vulnerability due to the complex interactions among the Indian, Eurasian, and Burmese tectonic plates. While recent advancements have employed Support Vector Machines (SVM), Convolutional Neural Networks (CNN), and Long Short-Term Memory (LSTM) for seismic prediction, these approaches often face challenges such as overfitting, limited generalization to unseen data, and difficulties in capturing non-linear temporal dependencies in complex geophysical signals. To address these limitations, this research employs a comprehensive dataset integrating various geological and climatic parameters specific to the Manipur region. Multiple machine learning classifiers, including ensemble methods and deep learning models, are applied to capture intricate precursory patterns indicative of seismic events. The proposed objective is to enhance prediction accuracy by identifying optimal model architectures and input features for earthquake forecasting. Among the evaluated models, the Random Forest algorithm demonstrates superior performance in handling non-linearities and feature interactions, achieving robust results across performance metrics such as Accuracy, Precision, Recall, F1-Score, and Area Under the Curve (AUC). A comparative analysis with traditional and state-of-the-art techniques confirms the efficacy of the proposed approach. Detailed observations on model behavior, along with challenges such as data noise, feature sparsity, and computational complexity, are discussed to provide deeper insight into the forecasting process.