Predicting Earthquake Location Using Convolutional Neural Network-Attention Mechanism Approach

In seismically active areas, earthquake prediction is essential for minimizing potential damages and preserving lives. However, precise forecasts are complicated to achieve because of seismic events’ complex and unpredictable nature. The current study presents an advanced prediction approach to address such issues, combining Convolutional Neural Networks (CNNs) and Attention Mechanism (AM). The primary goal is to improve the accuracy of the earthquake predictions and the generalizability across various mainland Chinese regions. AM layer emphasizes significant features for improving the prediction performance, whereas CNNs are utilized to extract spatial features of seismic data. The efficiency and effectiveness of the proposed approach were evaluated by comparing it with several well-known models. Results showed that the proposed approach performed consistently better than others in nine regions, with a reduced Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) as well as higher R-squared (R2) scores, especially in substantial seismic variability regions. Moreover, the proposed approach outperformed the conventional techniques in Region One, achieving an RMSE of 0.020, an MAE of 0.015, and an R2 value of 0.960. In regions susceptible to seismic events, this all-encompassing approach presents a promising path for earthquake prediction, boosting readiness and risk management methods.