[1]SHAO Lei,WANG Changzai.Application of the Density Clustering Algorithm in Machine Learning to Seismic Activity[J].Journal of Seismological Research,2025,(03):361-369.[doi:10.20015/j.cnki.ISSN1000-0666.2025.0038
]
Copy
Journal of Seismological Research[ISSN 1000-0666/CN 53-1062/P] Volume:
Number of periods:
2025 03
Page number:
361-369
Column:
Public date:
2025-04-15
- Title:
-
Application of the Density Clustering Algorithm in Machine Learning to Seismic Activity
- Author(s):
-
SHAO Lei1; WANG Changzai2
-
(1.Hunan Earthquake Agency,Changsha 410004,Hunan,China)(2.Institute of Geophysics,China Earthquake Administration,Beijing 100081,China)
-
- Keywords:
-
machine learning; the DBSCAN algorithm; seismic activity; precise earthquake location
- CLC:
-
P315.5
- DOI:
-
10.20015/j.cnki.ISSN1000-0666.2025.0038
- Abstract:
-
Aiming at the problem of effectively sub-dividing sub-regions for precise earthquake location in large-scale areas with intensive seismic activity,this paper,through multiple simulated data experiments and evaluations of various clustering algorithms in machine learning,concludes that the DBSCAN algorithm is most suitable for earthquake clustering in faults or earthquake swarms.Based on this,seismic activity studies are conducted to analyze and reveal spatial differences in the relative levels of current tectonic stress accumulation along fault zones,serving as an important foundation for analyzing the current activity patterns of faults(segments)and potential strong earthquake hazard zones.The DBSCAN algorithm was applied to cluster a total of 322 070 earthquake events in Sichuan-Yunnan region,resulting in nine distinct earthquake clusters.Among them,precise locations are determined using the double-difference earthquake location algorithm for the earthquake cluster in the intersection area of the southern segment of the Xianshuihe fault zone and the northern segment of the Anninghe fault zone,as well as for the Kangding MS6.3 earthquake sequence to the northwest,and the spatial distribution characteristics of earthquakes are also analyzed.The relocated results show that the Kangding earthquake sequence exhibits features consistent with the earthquake distribution on the strike-slip fault.The fault has an almost vertical dip angle.Earthquakes in the intersection area of the southern segment of the Xianshuihe fault zone and in the northern segment of the Anninghe fault zone are also densely distributed around the faults,with orientations generally consistent with the fault strikes.The longitudinal profile reveals the deep structural features of the fault.Furthermore,this paper investigates the spatial distribution characteristics of the b-value in the southern segment of the Xianshuihe fault zone and in the northern segment of the Anninghe fault zone from January 2009 to March 2021,revealing that the location of the Luding MS6.8 earthquake that occurred on September 5,2022 is in the area where b-value is obviously low.