Israel: New machine learning method helps predict earthquakes
News » Science & Technology
After the devastating earthquakes in Turkey, a team of researchers led by Yuval Reuveni from the Faculty of Physics at Ariel University published an article in the scientific journal Remote Sensing about a promising new way to predict aftershocks 48 hours earlier.
The revolutionary approach uses a form of machine learning called Support Vector Machine (SVM) with estimates of total electron content (TEC) in the GPS ionosphere.
The team has been studying the relationship between ionospheric electron content and geodynamic activity for several years. SVM is a set of machine learning algorithms that can be used to predict the occurrence of a certain event by analyzing the data set and identifying patterns and relationships.
Israeli scientists have found that the SVM algorithm accurately predicts seismic activity with up to 83% success – with an accuracy of 85.7% for true negative forecasts and 80% for true positive forecasts within 48 hours.
“Using machine learning approach, we were able to accurately predict earthquake events with a certain degree of success. While this method is not yet a reliable way to predict earthquakes, it is a promising step forward in our efforts to better understand seismic activity,” Reuveni said.
Using GPS receivers to evaluate TEC data, he said. is a cost-effective method that can be used to monitor geodynamic activity in real time. “This could potentially give people valuable time to prepare for an earthquake and also help mitigate the damage caused by seismic events,” the researcher emphasizes.
Follow us on Telegram