Machine learning in seismic structural design: an exploration of ANN and tabu-search optimization

Within the current seismology domain, earthquake magnitude prediction has become paramount since conventional approaches often need to improve precision and prognostic capability. This study discusses the urgent need for a prediction model that is more precise and dependable. The study presents a novel approach that utilizes sophisticated artificial neural networks (ANNs) and incorporates the tabu-search technique for hyperparameter tweaking to improve the model. The research employs a rigorous methodology using a comprehensive dataset that documents occurrences of earthquakes. The artificial neural network (ANN) model is trained across 50 epochs, with a batch size of 32. The key results demonstrate a significant R -squared value of 33.9%, indicating the improved predictive capacity of the model in estimating earthquake magnitudes. The mean absolute error (MAE) highlights its precision by exhibiting a variance of just 0.0806 units. The present study signifies a groundbreaking methodology for forecasting earthquake magnitudes, which has significant ramifications for seismic engineering and safety protocols.