UCSC and TU Munich Researchers Propose RECAST: A New Deep Learning-Based Model to Forecast Aftershocks

Artificial Intelligence finds its way into almost every possible field. There has been vast research going on in this domain. We are still a lot to discover. Artificial Intelligence and Deep Learning models also play an important role in Seismiography as they are used to predict earthquakes. For many previous years, the earthquake aftershock prediction models have stayed the same. These old models work fine with smaller datasets but struggle with bigger datasets. 

To fix this problem statement, researchers from the University of California, Santa Cruz, and the Technical University of Munich made a new model that uses Deep Learning called RECAST. They used Deep Learning behind this model, as it is useful for handling larger datasets. The new model was effective compared to the older model as it defeated the old one in every possible way. The old earthquake prediction model, ETAS was created a few years ago when these researchers had limited data. But today, we have huge datasets, which the old model couldn’t work on. The old ETAS model is fragile and tricky to use. To improve earthquake prediction with deep learning, we need a better way to compare models. The RECAST model was tested with both synthetic and real earthquake data from Southern California. It performed slightly better than the ETAS model, especially with more data, and it was faster, too.

Researchers have tried using Machine Learning and Deep Learning models to predict earthquakes before, but the technology wasn’t quite ready. The RECAST model is more accurate and can easily work with different earthquake datasets. This flexibility could revolutionize earthquake forecasting. With deep learning, models can handle lots of new data and even combine information from various regions to predict earthquakes in less-studied areas. This information about the Deep Learning models was quite useful and was being researched. Researchers also examined that a model trained on New Zealand, Japan, and California data could be used to forecast earthquakes in places with less available data.

These Deep Learning models will also help researchers access different data types for earthquake prediction. They can now use continuous ground motion data instead of focusing on something officially classified as an earthquake. This is a classification task. The model’s accuracy and F1 score were quite good for the larger datasets. The researchers are still working on this new model that will encourage and motivate discussions about all the possibilities because it has a lot of potential to do.


Check out the Paper and Reference Article. All Credit For This Research Goes To the Researchers on This Project. Also, don’t forget to join our 30k+ ML SubReddit, 40k+ Facebook Community, Discord Channel, and Email Newsletter, where we share the latest AI research news, cool AI projects, and more.

If you like our work, you will love our newsletter..

Bhoumik Mhatre is a Third year UG student at IIT Kharagpur pursuing B.tech + M.Tech program in Mining Engineering and minor in economics. He is a Data Enthusiast. He is currently possessing a research internship at National University of Singapore. He is also a partner at Digiaxx Company. 'I am fascinated about the recent developments in the field of Data Science and would like to research about them.'

🐝 Join the Fastest Growing AI Research Newsletter Read by Researchers from Google + NVIDIA + Meta + Stanford + MIT + Microsoft and many others...