Earthquake prediction model using support vector regressor and hybrid neural networks.

Area of Science:

• Geophysics
• Computational Seismology

Background:

• Earthquake prediction remains a significant challenge in seismology.
• Accurate prediction is crucial for mitigating catastrophic impacts.

Purpose of the Study:

• To develop an improved earthquake prediction system.
• To evaluate the system's performance in diverse seismically active regions.

Main Methods:

• Computed sixty seismic features using seismological concepts like the Gutenberg-Richter law.
• Applied Maximum Relevance and Minimum Redundancy (mRMR) for feature selection.
• Developed a Support Vector Regressor (SVR) and Hybrid Neural Network (HNN) model with Enhanced Particle Swarm Optimization (EPSO) for weight optimization.

Main Results:

• The SVR-HNN system, utilizing newly computed seismic features, was applied to the Hindukush, Chile, and Southern California regions.
• Numerical results indicated enhanced prediction performance across all tested regions.
• The proposed method outperformed previous earthquake prediction studies.

Conclusions:

• The integration of advanced seismic features with an SVR-HNN prediction system offers a promising advancement in earthquake forecasting.
• The developed methodology shows significant potential for improving the accuracy and reliability of earthquake predictions globally.