Nur Hasanah, Zhihong Man, Zhenwei Cao, Jinchuan Zheng
In this work, a nonlinear autoregressive exogenous neural network (NARXNN) is combined with the Cubature Kalman Filter (CKF) algorithm to predict the earthquakes in Java Island, Indonesia. It is seen that CKF when incorporated into NARXNN, can have positive effects on the estimation of state and prediction accuracy as well as the ability to handle nonlinear dynamics. The NARXNN utilizes cubature integration and a collection of sigma points to effectively capture and predict its nonlinear characteristics. The proposed model was then used to predict the next 15 earthquakes in Java. In comparison to a regular NARXNN model, the trained neural model produced smaller Mean Squared Error and highly accurate predictions. The results showed that the proposed model was able to learn from complex seismic patterns and capture the nonlinear characteristics of Java. This demonstrates the potential of the proposed model to predict future earthquakes in other areas. © 2023 IEEE.
Swinburne University of Technology, Department of Engineering Technologies, Melbourne, Australia; Universitas Negeri Yogyakarta, Department of Electronics and Informatics Education, Yogyakarta, Indonesia
