An Innovative BiLSTM-Based Method for Earthqu...

An Innovative BiLSTM-Based Method for Earthquake Prediction System

Author Name : Aditya Maheshwari, Udit Sathe

DOI: https://doi.org/10.56025/IJARESM.2025.1311250030

ABSTRACT This paper presents an innovative bidirectional long short-term memory (BiLSTM) based method for earthquake prediction that integrates advanced deep learning techniques to achieve superior performance in seismic forecasting. The proposed system combines convolutional neural networks (CNN) for spatial feature extraction, BiLSTM networks for temporal dependency modeling, attention mechanisms for feature importance weighting, and a novel LUTanh activation function for enhanced learning capability. Through comprehensive analysis of three seminal research papers and extensive experimentation on multi-regional seismic datasets, our method demonstrates significant improvements over existing approaches, achieving 84.7% accuracy, 0.42 MAE, and 0.891 R² score [1][2][3]. The system successfully predicts earthquake occurrence, magnitude, location, and depth with high precision, making it suitable for real-world seismic monitoring and early warning applications.