使用支持向量机识别地震类型的影响因素分析  

Analysis of Influencing Factors in Use of Support Vector Machine Method to Identify Earthquake Types

作　　者：范晓易 曲均浩[2] 刘方斌[2] 周少辉 FAN Xiaoyi;QU Junhao;LIU Fangbin;ZHOU Shaohui(Tai’an Fiducial Seismic Station of Shandong Earthquake Agency,2 Luohanya Road,Tai’an 271000,China;Shandong Earthquake Agency,20 East-Wenhua Road,Jinan 250014,China)

机构地区：[1]山东省地震局泰安基准地震台,山东省泰安市271000 [2]山东省地震局,济南市250014

出　　处：《大地测量与地球动力学》2020年第10期1034-1038,共5页Journal of Geodesy and Geodynamics

基　　金：中国地震局地震科技星火计划(XH19027);山东省地震局科研项目(YB2003)。

摘　　要：对山东地区2006~2017年3种地震事件——天然地震、爆破及塌陷的波形记录进行小波变换,对提取出的香农熵特征采用支持向量机LIBSVM方法进行分类识别,并设计一系列实验研究影响最终分类效果的因素。结果表明,5种影响因素——信号窗长度、小波分解方式、小波基类型、向量机算法类型、向量机核函数类型均对地震类型的分类识别结果产生一定的影响;识别率最高的3组处理方式均采用了2000 s信号窗长度+db7小波基+υ-SVC算法的组合方式。所得的识别率较高的几种影响因素组合,在未来可应用于地震类型的实时识别,进一步提高地震类型的识别率和触发准确率。In this paper,the authors used the waveform records of the 2006 to 2017 earthquake events in Shandong province,and the wavelet transform of three kinds of earthquake type:natural seismic,blasting and collapse waveform are carried out,and Shannon entropy features are extracted by support vector classifier LIBSVM.A series of experiments are designed to study the factors that affect the final classification effect.The results show that the length of signal window,the way of wavelet decomposition,the type of wavelet base,the type of vector machine and the type of vector machine kernel function all have some influence on the result of seismic classification.The combination of 2000 seconds signal window length+db7 wavelet base+υ-SVC vector machine is used in the three groups with the highest recognition rate.The combination of several factors with high recognition rate can be applied to real-time recognition of earthquake type in the future to further improve the recognition rate of earthquake type and trigger accuracy.

关 键 词：支持向量分类机 香农熵 特征向量 分类识别 影响因素 

分 类 号：P315[天文地球—地震学]