Seismicity analysis and machine learning models for short-term low magnitude seismic activity predictions in Cyprus

Effective management and planning for the sustainable development of urban regions requires a wide range of up-to-date and impartial information. This study focusses on earthquake catalog-based seismicity analysis for Cyprus region. It is followed by computation of seismic features and short-term prediction of seismic activity using machine learning techniques. Earthquake catalog is investigated temporally and noisy data is removed. Sixty seismic features were then computed based upon cleaned earthquake catalog to express the internal seismic state of the region. These seismic features are then modeled using machine learning techniques with the corresponding seismic activity. Three machine learning algorithms, namely Artificial Neural Networks, Support Vector Machines and Random Forests, are employed for seismic activity prediction. The framework is designed to obtain five days-ahead, one week-ahead, ten days-ahead and fifteen days-ahead predictions for moment magnitude thresholds of 3.0, 3.5, 4.0 and 4.5. Based on the Matthews correlation coefficient (MCC), the predictions obtained using the Random Forest were found to be the most accurate for magnitude thresholds of 3.0 and 3.5 across all the prediction periods. Similarly, the predictions obtained using the Support Vector Machine outperformed other techniques for magnitude thresholds of 4.0 and 4.5.