利用地震震级样本数据验证PCA-GSM-GRNN模型的优越性  

Using the Data of Earthquake Magnitude Samples to Verify the Superiority of PCA-GSM-GRNN Model

作　　者：王晨晖 常玉柱 袁颖 王秀敏 WANG Chenhui;CHANG Yuzhu;YUAN Ying;WANG Xiumin(National Field Scientific Observation and Research Station for Huge Thick Sediments and Seismic Disasters in Hongshan,Hebei Province,Hebei Xingtai 054000;Xingtai Central Seismic Station of Hebei Earthquake Agency,Hebei Xingtai 054000,China;Chengde Central Seismic Station of Hebei Earthquake Agency,Hebei Chengde 067000,China;School of Urban Geology and Engineering,Hebei Geologic University,Hebei Shijiazhuang 050031;Hebei Province Underground Artificial Environment Intelligent Development and Control Technology Innovation Center,Hebei Shijiazhuang 050031)

机构地区：[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 [2]河北省地震局邢台地震监测中心站,河北邢台054000 [3]河北省地震局承德地震监测中心站,河北承德067000 [4]河北地质大学城市地质与工程学院,河北石家庄050031 [5]河北省地下人工环境智慧开发与管控技术创新中心,河北石家庄050031

出　　处：《四川地震》2024年第2期35-38,共4页Earthquake Research in Sichuan

基　　金：国家自然科学基金(41807231);河北地质大学科技创新团队项目(KJCXTD-2021-08);河北省地震科技星火计划项目(DZ2021110500001)

摘　　要：针对地震震级与其影响指标之间的非线性问题,提出了基于网格搜索法(GSM)和主成分分析法(PCA)优化广义回归神经网络(GRNN)的地震预测模型。采用PCA对震级影响指标进行维度约简,将降维后的主成分作为模型输入向量,地震震级作为模型输出向量,同时选用GSM寻优GRNN最佳参数,利用学习样本对新模型进行训练,最终构建基于PCA-GSM-GRNN的地震震级预测模型。将PCA-GSM-GRNN模型应用于测试样本,结果显示:PCA-GSM-GRNN模型预测结果准确率相较于GRNN-GSM和GRNN模型分别提高5.03%和5.66%,具有良好的预测效果。We proposed the earthquake prediction model that is based on the general regression neural network(GRNN)optimized by the grid search method(GSM)and principal component analysis(PCA),which addressed the nonlinear relationship between earthquake magnitude and its impact indicators.Moreover,we used PCA to reduce the dimensionality of the impact indicators and applied the reduced principal components as the input vectors of the model,which required the earthquake magnitude as the output vector.We further used GSM to optimize the best parameters of the GRNN,and trained the model by learning samples to construct the earthquake magnitude prediction model based on PCA-GSM-GRNN.Finally,we applied the PCA-GSM-GRNN model to the test samples.The results showed that the accuracy of the PCA-GSM-GRNN model prediction was improved by 5.03%and 5.66%compared with the GSM-GRNN model and GRNN model,which indicates a good prediction performance.

关 键 词：地震震级 主成分分析法 网格搜索法 广义回归神经网络 

分 类 号：P315.7[天文地球—地震学]