Abstract:

Being able to predict earthquakes has been an important issue for human life and comfort. However, earthquakes with a complex mechanism are quite difficult to predict. While different methods were used to predict earthquakes in the past, these methods have been used to make earthquake predictions with the developments in artificial intelligence methods recently. In this study, the focal coordinates and focal depths of earthquakes with a magnitude of 6 and above, which are likely to occur in the northwestern region of Turkey, were tried to be estimated. In this study, six different machine learning methods (Support Vector Machines, Linear Regression, Gradient Boost, Elastic Net, Bayesian Ridge and XGBoost) were used comparatively and the prediction results were compared. Results were evaluated with RMSE, MAE, and Adjusted R2 performance metrics. The estimation results show that earthquakes of 6 or more may occur in the future in the northwestern region of Turkey, within the borders of Bursa province, starting from the north of Lake Iznik and going westwards in Ekinli, northwest of İmralı island, north of Avşa island and northwest of Marmara island.

Keywords: