Monthly Maximum Magnitude Prediction in the North–South Seismic Belt of China Based on Deep Learning DOI Creative Commons
Ning Mao, Ke Sun, Jingye Zhang

et al.

Applied Sciences, Journal Year: 2024, Volume and Issue: 14(19), P. 9001 - 9001

Published: Oct. 6, 2024

The North–South Seismic Belt is one of the major regions in China where strong earthquakes frequently occur. Predicting monthly maximum magnitude significant importance for proactive seismic hazard defense. This paper uses catalog data from since 1970 to calculate and extract multiple parameters. processed using Variational Mode Decomposition (VMD) with sample segmentation avoid information leakage. decomposed modal parameters together form a new dataset. Based on these datasets, this employs four deep learning models time windows predict magnitude, prediction accuracy (PA), False Alarm Rate (FAR), Missed (MR) as evaluation metrics. It found that window 12 generally yields better results, PA Ms 5.0–6.0 reaching 77.27% above 6.0 12.5%. Compared not VMD, traditional error metrics show only slight improvement, but model can short-term trends changes.

Language: Английский

Monthly Maximum Magnitude Prediction in the North–South Seismic Belt of China Based on Deep Learning DOI Creative Commons
Ning Mao, Ke Sun, Jingye Zhang

et al.

Applied Sciences, Journal Year: 2024, Volume and Issue: 14(19), P. 9001 - 9001

Published: Oct. 6, 2024

The North–South Seismic Belt is one of the major regions in China where strong earthquakes frequently occur. Predicting monthly maximum magnitude significant importance for proactive seismic hazard defense. This paper uses catalog data from since 1970 to calculate and extract multiple parameters. processed using Variational Mode Decomposition (VMD) with sample segmentation avoid information leakage. decomposed modal parameters together form a new dataset. Based on these datasets, this employs four deep learning models time windows predict magnitude, prediction accuracy (PA), False Alarm Rate (FAR), Missed (MR) as evaluation metrics. It found that window 12 generally yields better results, PA Ms 5.0–6.0 reaching 77.27% above 6.0 12.5%. Compared not VMD, traditional error metrics show only slight improvement, but model can short-term trends changes.

Language: Английский

Citations

0