Integration of Fuzzy System with Convolutional Neural Networks and Bidirectional Long Short-Term Memory for Earthquake Prediction

Earthquake prediction is one of the critical research area addressing significant challenges that are associated with disaster preparedness and public safety. The uncertainty related to earthquake timing, location, and details is the primary problem, usually leading to responses that are inadequate. The aim of the study is enhancing the accuracy of predictions and mitigating uncertainty through the development of a novel model with the use of advanced ML techniques. The suggested approach uses Bidirectional Long Short-Term Memory (BiLSTM) networks for temporal data analysis and Convolutional Neural Networks (CNN) for spatial data analysis, augmented by an Attention Mechanism in order to prioritize important features as well as fuzzy system to improve uncertainty. An 11,442-earthquake dataset, data pre-processing methods have been utilized for enhancing model training and performance. The results showed an increase in prediction accuracy, obtaining a 0.015 from the Mean Absolute Error (MAE), a lower value of 0.020 from the Root Mean Squared Error (RMSE), and 0.880 from the R2, a significant value. This indicates that the system explains 88% of the variance. This demonstrates the system’s effectiveness in dealing with uncertainty and providing important data for risk management and prediction of earthquakes.