基于SVM的地震序列类型早期预测研究*
李冬梅1,2,周翠英2,朱成林2,孙龙梅3,吕子强2,程显洲1

(1.中国地震局兰州地震研究所,甘肃 兰州 730000; 2.山东省地震局,山东 济南 250014; 3.山东女子学院,山东 济南 250300)

地震序列; SVM; 统计模式识别; 序列类型; 早期预测

Research on Early Judgment of Earthquake Sequence Types Based on SVM
LI Dong-mei1,2,ZHOU Cui-ying2,ZHU Cheng-lin2,SUN Long-mei3,

(1.Lanzhou Institute of Seismology,CEA,Lanzhou 730000,Gansu,China)(2. Earthquake Administration of Shandong Province,Jinan 250014,Shandong,China)(3.Shandong Women's University,Jinan 250300,Shandong,China)

seismic sequence; SVM; statistic pattern recognition; sequence type; early prediction

备注

在Matlab环境下,通过构造SVM,建立地震序列特征参数与序列类型之间的一种非线性映射关系对地震序列类型进行早期分类预测。依据我国1970年以来的MS≥5.0地震序列资料,使用SVM对震后1、2、3、5、7天5个时间尺度的地震序列类型进行早期预测,识别效果较好,处理速度快,具有较强的实用性。

In the environment of Matlab, we construct Support Vector Machine(SVM)to build a kind of nonlinear mapping relationship between seismic sequence characteristic parameter and sequence type, and do the early predictions of earthquake sequence types. On the basis of MS≥5.0 earthquake sequences in China since 1970, we divide the data in 5 time scales according to one, two, three, five and seven days after the earthquake and apply the SVM to the early predictions of earthquake sequence types. The results show that it achieves good recognition and fast processing speed, and has a strong practicability.