Performance Analysis of Machine Learning and Deep Learning Architecture for Earthquake Magnitude Prediction

Earthquakes possess immense destructive power, capable of causing significant harm to buildings and the natural surroundings. Accurate earthquake magnitude prediction is crucial for determining seismic risks and implementing appropriate mitigation measures. Although deep learning (DL) and machine learning (ML) algorithms have been utilized extensively in earth-quake prediction research, a thorough DL vs. ML comparison is absent. The comparison is broadened in this research by concentrating on earthquake magnitude prediction. Moreover, we employ a set of innovative algorithms that have had limited application within this domain. The optimal algorithm to estimate the magnitude of an earthquake has not been explicitly discussed in any of the papers. This study aims to bridge this gap by employing a diverse range of twelve algorithms, i.e. Decision Tree Regressor, Random Forest Regressor, SVR, Linear Regressor, Gradient Boosting, XGBoost, LightGBM, CatBoost, ANN, RNN, LSTM, and MLP, encompassing both machine learning and deep learning techniques, to predict earthquake magnitudes. Through comprehensive experimentation and comparative research, we evaluate the performance of the ensemble-based models in terms of different evaluation metrics, such as R
²
Scores, MAE, MSE and RMSE.