Machine learning model achieves 97.97% accuracy in Los Angeles earthquake predictions, California

Researchers from the Department of Information Technology at Georgia Southern University have introduced a new foundation for earthquake prediction by developing advanced machine learning and neural network models built with a detailed feature matrix to maximize predictive accuracy in Los Angeles, California.

They believe that applying a carefully designed feature set to the Random Forest machine learning model provides accurate predictions of the maximum earthquake category over the next 30 days. Compared to the other 16 machine learning algorithms tested, Random Forest proved to be the most effective.

Professor Lei Chen, co-author of this research from the Department of Information Technology at Georgia Southern University, confirmed that this research opens new doors for applying machine learning to disaster risk management, offering predictive tools that can help authorities prepare for risks and hazards.

“The integration of advanced machine learning algorithms like Random Forest and neural networks has allowed us to break new ground in seismic forecasting,” co-author of the study, Professor Yiming Ji, also from the Department of Information Technology at Georgia Southern University, added,

“Our team’s work not only pushes the boundaries of earthquake prediction but also sets the stage for future advancements in applying machine learning to other natural disaster forecasting models. The implications for improving public safety and emergency response are vast,” said Professor Christopher Kadlec, Associate Professor at Georgia Southern University.

With a 69.14% accuracy rate in predicting the maximum earthquake magnitude within one of six categories, researchers created a predictive pattern matrix for Los Angeles in previous studies.

They expanded their research to Istanbul, Turkey, one of the most earthquake-prone zones, and achieved an accuracy rate of 91.65%. Further development brought the accuracy rate to 98.53% for San Diego.

Following successful results in San Diego and Istanbul, the researchers returned to Los Angeles to improve upon the previous 69.14% accuracy rate, achieving 97.97% this time.

They state that these results suggest that machine learning techniques may significantly enhance earthquake prediction accuracy, providing authorities with a more effective means to prepare for risks and hazards.

“Our model’s 97.97% accuracy marks a significant improvement over traditional methods, offering critical insights that can save lives and reduce property damage in high-risk areas,” said Cemil Emre Yavas, another co-author of the study.

Researchers used a combination of machine learning and neural network techniques to predict seismic activities in Los Angeles, drawing on a dataset covering earthquake reports from the past 12 years. Advanced feature engineering enabled them to create a matrix incorporating crucial predictive inputs.

The scientists used earthquake data from the Southern California Earthquake Data Center (SCEDC), managed by the California Institute of Technology.

Earthquake prediction accuracy can be improved by detecting deep seismic patterns, testing multiple tools, and investigating seismic frequency features. Using this study as a foundation, scientists created and tested 16 different machine learning and neural network algorithms to select the most effective model for predicting the largest earthquake magnitude within 30 days.

This research aims to enhance predictive modeling tools for the Los Angeles region by integrating findings from other earthquake prediction studies. The researchers seek to increase earthquake forecast accuracy through the combination of machine learning algorithms, feature extraction methods, and advanced neural network topologies.

A 100 km (62 miles) radius was chosen to encompass a broad area around Los Angeles highly relevant to earthquake forecasting.

Using a radius of less than 100 km (62 miles) may exclude key seismic events crucial to understanding earthquake patterns in the region.

Conversely, a radius larger than 100 km (62 miles) could introduce noise by including data from areas with distinct seismic properties, potentially decreasing the model’s forecasting accuracy.

Thus, a 100 km (62 miles) radius strikes an ideal balance, ensuring sufficient data while maintaining the model’s relevance and accuracy.

When fine-tuned with suitable hyperparameters, the Random Forest model yields robust and accurate predictions. These hyperparameters allow the model to utilize the full complexity of the data, thereby enhancing its predictive capabilities.

This fine-tuned model can support further analysis and could serve as a valuable tool in earthquake forecasting.

The research builds upon an extensive array of earthquake prediction studies conducted between 1990 and 2024.

Advanced neural network models, such as the graph convolutional neural network described by Bilal et al., can greatly enhance earthquake prediction performance. Their focus on early earthquake detection using complex neural network designs demonstrates the potential of advanced technologies to improve predictive capabilities.

Initiatives such as the Collaboratory for the Study of Earthquake Predictability (CSEP) and the Regional Earthquake Likelihood Models Experiment (RELM) by Schorlemmer et al. (2010) have paved the way for potential earthquake prediction advancements.

Research drawing on diverse data sources, such as GPS, ionospheric data, and outgoing longwave radiation, has improved earthquake forecasting models. Gitis et al. (2021) showed the value of seismological data, while Zhai et al. (2020) explored thermal anomalies using non-seismic time series data, illustrating the multidisciplinary nature of earthquake research.

The work of Hsu and Pratomo (2022) on early peak ground acceleration prediction using Long Short-Term Memory (LSTM) neural networks demonstrated the utility of models that capture order dependency in seismic waves. This aligns with the approach of employing machine learning techniques to calculate earthquake occurrences over a specified timeframe.

References:

¹ Improving earthquake prediction accuracy in Los Angeles with machine learning – Cemil Emre Yavas et al. – Nature Scientific Reports – October 18, 2024 – https://doi.org/10.1038/s41598-024-76483-x – OPEN ACCESS

TW

The Watchers team and our contributors bring the latest on extreme weather, earthquakes, volcanic eruptions, space weather, and all things science. We're all about making sense of the natural world and keeping you informed on what’s happening. Got a tip or a question? Hit us up using the form at newstips!

• TW

  DNA study reshapes understanding of Pompeii’s ancient population
• TW

  Severe Heatwave Warning over much of Queensland, Australia
• TW

  Floodwaters wash away 30 vehicles in Cadaqués, Catalonia, Spain
• TW

  93 still missing after severe floods in Spain