Earthquake Prediction In Turkey With Structural Recurrent Neural Networks

Although the studies for the prediction of earthquakes encourage people from time to time, it is generally considered that there has been no success in earthquake prediction. Studies in earthquake prediction, generally focus on the temporal interaction of earthquakes in certain regions. In this study, predictions are generated with using structural recurrent neural networks (SRNN), which deals with the temporal and spatial patterns of earthquakes as a whole. In 9 equal regions of Turkey within a month, whether there will be an earthquake of magnitude 4 or above is predicted with 0.72 accurracy and 0.74 average precision.