A Model Fusion of 1D CNN with NARX for Accurate Earthquake Time Series Prediction

we proposed a fusion model for one-dimensional convolutional neural networks (1D-CNN) with nonlinear autoregressive with exogenous inputs (NARX) that offers a potential method to improve the accuracy of predictions. The objective of this project is to create and assess a fusion 1D-CNN-NARX model for precise forecasting of earthquake time series. The data set consists of seismic records from West Java, Central Java, and East Java. The results indicate that the fusion 1D-CNN-NARX model achieves a high level of accuracy in predicting future outcomes, especially for short-term forecasts. The mean squared error (MSE) values for West Java, Central Java, and East Java are 9.3590e-06, 7.2549e-06, and 2.6432e-05, respectively. The results show that model is effective for short-term earthquake prediction in all regions, with accuracies generally exceeding 95% for the first three months. This study emphasizes the possibility of integrating 1D-CNN and NARX architectures to enhance earthquake prediction approaches, providing a reliable and precise tool for forecasting seismic activity and early warning systems.