GSM-SVM在地震震级预测中的应用
APPLICATION OF GSM-SVM IN EARTHQUAKE MAGNITUDE PREDICTION
王晨晖 1吕国军 2王秀敏 1畅国平1
作者信息
- 1. 河北红山巨厚沉积与地震灾害国家野外科学观测研究站,河北邢台 054000;河北省地震局邢台地震监测中心站,河北邢台 054000
- 2. 河北红山巨厚沉积与地震灾害国家野外科学观测研究站,河北邢台 054000;河北省地震局,河北石家庄 050031
- 折叠
摘要
针对地震震级影响因子众多且关系重复等问题,为合理预测地震震级,提出了基于网格搜索法优化支持向量机(support vector machine,SVM)的地震震级预测模型.选取地震累积频度、累积释放能量、b值、异常震群个数、地震条带个数、活动周期和相关区震级等7个影响因子,利用主成分分析法(principal component analysis,PCA)去除因子间的冗余信息,降低输入维数,并利用网格搜索法(grid search method,GSM)确定SVM参数C和g,建立震级预测模型,并对测试样本进行预测,与遗传算法(genetic algorithm,GA)和粒子群算法(particle swarm optimization,PSO)预测结果相对比,结果表明:PCA-GSM-SVM模型预测结果平均相对误差为1.29％,具有较高的预测精度.
Abstract
In view of numerous and redundant factors affecting earthquake magnitude,in order to predict earthquake magnitude effectively,a seismic magnitude prediction model based on support vector machine(SVM)optimized by grid search method was proposed.Selecting 7 influencing fac-tors,including cumulative earthquake frequency,cumulative released energy,b-value,number of abnormal earthquake swarms,number of seismic bands,activity period,and magnitude of related areas.Principal component analysis(PCA)was used to remove redundant information between fac-tors,and the input dimensions was reduced,and grid search method(GSM)was used to determine SVM parameters C and g,finally the magnitude prediction model was established,which was used to predict test samples,and compared with the prediction results of Genetic Algorithm(GA)and Parti-cle Swarm Optimization(PSO).The results showed that the average relative error of PCA-GSM-SVM was 1.29％,which had higher prediction accuracy.
关键词
GSM-SVM/地震震级预测/主成分分析法/网格搜索法/支持向量机Key words
Earthquake magnitude/Principal component analysis/Grid search method/Support vector machine引用本文复制引用
基金项目
河北省地震科技星火计划(DZ2021121600001)
河北省地震科技星火计划(DZ2023120800022)
河北省重点研发计划(22375406D)
出版年
2024
NETL
NSTL
万方数据