Comprehensive index analyses on seismic activity parameters based on data mining for MS≥5.0 earthquakes in northeast China and North China
-
摘要: 根据地震活动参数的数据特征,以2013年内蒙古科尔沁MS5.3地震为例,引入主成分分析和因子分析两种数据挖掘方法,在信息损失尽可能少的条件下,实现对参数变量的约简降维,并提取综合指标W.研究显示,2013年科尔沁MS5.3地震前两年,W值变化显著.在进一步的单因子分析中,从综合后的参数变量信息中重新提取了具有物理意义的多个单因子综合指标,消除了大量信息重叠所造成的不一致,实现了对综合指标的细化分析.在此基础上,对东北华北29次MS≥5.0地震的11个地震活动参数(频次N,b值,η值,A(b)值,Mf值,AC值,C值,D值,E值,Rm值,响应比Y)进行主成分分析,其结果显示,主成分综合指标W在震前1-2年均出现了明显的异常变化,这充分说明综合指标W可以用作地震预报研究的综合异常参考判据.Abstract: According to the data characteristics of seismic activity parameters, taking the MS5.3 Horqin, Inner Mongolia earthquake in 2013 for a typical case, the present paper introduces principal component analysis method and factor analysis method to reduce the dimension of parameter variables on the condition of information loss as little as possible, and extracts the comprehensive index. The result shows that the comprehensive index W had been changed significantly two years before the earthquake. Furthermore, by conducting factor analysis on Horqin earthquake, several single factor comprehensive indices with physical meanings are extracted, not only eliminating the inconsistency caused by the overlapping of information, but also achieving the refinement of the comprehensive index. In further research, the principal component analysis result about the eleven seismic activity parameters (N, b, η, A(b), Mf, AC, C, D, E, Rm, Y) of 29 earthquakes with MS≥5.0 in northeast China and North China shows that comprehensive index W of principal component analysis had obvious precursor changes in 1-2 years before the earthquakes. This suggests that the index W can be taken as a precursory in earthquake prediction research.
-
-
表 1 2013年科尔沁MS5.3地震11个活动性参数的相关性系数
Table 1 Correlativity between the 11 seismic activity parameters of 2013 HorqinMS5.3 earthquake
N b η A(b) Mf AC C D E Rm Y N 1.000 -0.201 0.205 0.497 0.395 0.184 0.882 0.850 0.635 0.169 -0.047 b -0.201 1.000 -0.010 -0.299 -0.389 -0.146 -0.167 -0.244 -0.245 -0.230 0.043 η 0.205 -0.010 1.000 0.805 0.067 0.574 0.177 0.259 0.268 0.294 -0.097 A(b) 0.497 -0.299 0.805 1.000 0.533 0.713 0.438 0.540 0.386 0.388 -0.078 Mf 0.395 -0.389 0.067 0.533 1.000 0.465 0.391 0.403 0.114 0.187 0.018 AC 0.184 -0.146 0.574 0.713 0.465 1.000 0.176 0.200 -0.117 0.419 -0.005 C 0.882 -0.167 0.177 0.438 0.391 0.176 1.000 0.670 0.473 0.141 0.064 D 0.850 -0.244 0.259 0.540 0.403 0.200 0.670 1.000 0.606 0.247 -0.139 E 0.635 -0.245 0.268 0.386 0.114 -0.117 0.473 0.606 1.000 0.021 -0.141 Rm 0.169 -0.230 0.294 0.388 0.187 0.419 0.141 0.247 0.021 1.000 0.035 Y -0.047 0.043 -0.097 -0.078 0.018 -0.005 0.064 -0.139 -0.141 0.035 1.000 表 2 2013年科尔沁MS5.3地震活动参数主成分特征值和贡献率
Table 2 Eigenvalues and contribution rate of Horqin MS5.3 earthquake in 2013
主成分 特征值 贡献率 累计贡献率 1 4.331 39.371% 39.371% 2 1.920 17.451% 56.822% 3 1.232 11.201% 68.023% 4 1.049 9.532% 77.555% 5 0.820 7.457% 85.012% 6 0.726 6.600% 91.613% 7 0.370 3.366% 94.979% 8 0.257 2.335% 97.314% 9 0.199 1.809% 99.123% 10 0.054 0.488% 99.611% 11 0.043 0.389% 100.000% 表 3 2013年科尔沁MS5.3地震活动参数主成分载荷系数
Table 3 Load coefficient of principal component of Horqin MS5.3 earthquake in 2013
参数 主成分载荷系数 1 2 3 4 5 N 0.832 -0.461 -0.003 0.157 -0.059 b -0.395 -0.009 0.499 0.570 -0.326 η 0.558 0.512 0.544 0.122 0.095 A(b) 0.859 0.400 0.151 0.002 -0.049 Mf 0.600 0.137 -0.546 -0.146 -0.422 AC 0.547 0.718 -0.053 0.075 -0.228 C 0.748 -0.407 -0.103 0.285 -0.112 D 0.823 -0.354 0.048 -0.004 0.006 E 0.596 -0.514 0.300 -0.111 0.270 Rm 0.417 0.438 -0.151 -0.009 0.555 Y -0.092 0.055 -0.486 0.751 0.275 表 4 2013年科尔沁MS5.3地震因子载荷系数
Table 4 Load coefficient of factors of Horqin MS5.3 earthquake in 2013
参数 因子载荷系数 1 2 3 4 N 0.940 0.136 0.153 0.059 b -0.122 -0.050 -0.822 0.190 η 0.195 0.853 -0.272 -0.210 A(b) 0.418 0.837 0.187 -0.101 Mf 0.271 0.312 0.705 0.179 AC -0.025 0.871 0.227 0.113 C 0.854 0.125 0.138 0.229 D 0.844 0.205 0.205 -0.093 E 0.784 -0.008 -0.016 -0.326 Rm 0.022 0.546 0.287 0.089 Y -0.026 -0.020 -0.055 0.898 -
国家地震局预测预防司. 1997.测震学分析预报方法[M].北京:地震出版社: 49-116. Department of Prediction and Prevention, State Seismological Bureau. 1997. Analysis and Prediction Methods of Seismometry[M]. Beijing: Seismological Press: 49-116 (in Chinese).
韩渭宾. 2003.地震活动性参数分类及其相关性初步研究[J].四川地震, (3): 1-5. http://www.cnki.com.cn/Article/CJFDTOTAL-SCHZ200303001.htm Han W B. 2003. Primary study on the classifying among the seismicity parameters and their correlativity[J]. Earthquake Research in Sichuan, (3): 1-5 (in Chinese). http://www.cnki.com.cn/Article/CJFDTOTAL-SCHZ200303001.htm
华爱军, 刘西林, 刁守中, 李红. 2001.华北地区中等、大地震前算法复杂性AC值异常变化特征[J].内陆地震, 15 (4): 326-330. Hua A J, Liu X L, Diao S Z, Li H. 2001. Abnormal characteristics of arithmetic complexity AC value before moderate, strong earthquakes in North China[J]. Inland Earthquake, 15 (4): 326-330 (in Chinese).
李永振. 2011.主成分分析法在辽宁地区地震预测中的应用[J].西北地震学报, 33 (1): 76-79. http://www.cnki.com.cn/Article/CJFDTOTAL-ZBDZ201101015.htm Li Y Z. 2011. Application of primary components analysis method to earthquake prediction in Liaoning Province[J]. Northwestern Seismological Journal, 33 (1): 76-79 (in Chinese). http://www.cnki.com.cn/Article/CJFDTOTAL-ZBDZ201101015.htm
陆远忠, 阎利军, 郭若眉. 1999.用于中短期地震预报的一些地震活动性参量相关性讨论[J].地震, 19 (1): 11-18. http://www.cnki.com.cn/Article/CJFDTOTAL-DIZN199901001.htm Lu Y Z, Yan L J, Guo R M. 1999. Discussion about correlation of some seismological parameters in medium-and short-term earthquake prediction[J]. Earthquake, 19 (1): 11-18 (in Chinese). http://www.cnki.com.cn/Article/CJFDTOTAL-DIZN199901001.htm
王炜, 林命週, 赵利飞, 马钦忠. 2006a.地震活动参数约简的因子分析方法[J].西北地震学报, 28 (4): 303-308. http://www.cnki.com.cn/Article/CJFDTOTAL-ZBDZ200604002.htm Wang W, Lin M Z, Zhao L F, Ma Q Z. 2006a. Factor analysis method for reducing seismicity parameters[J]. Northwestern Seismological Journal, 28 (4): 303-308 (in Chinese). http://www.cnki.com.cn/Article/CJFDTOTAL-ZBDZ200604002.htm
王炜, 刘悦, 李国正, 吴耿锋, 马钦忠, 赵利飞, 林命週. 2006b.我国大陆强震预测的支持向量机方法[J].地震学报, 28 (1): 29-36. http://www.dzxb.org/Magazine/Show?id=26517 Wang W, Liu Y, Li G Z, Wu G F, Ma Q Z, Zhao L F, Lin M Z. 2006b. Support Vector Machine method for forecasting future strong earthquakes in Chinese mainland[J]. Acta Seismologica Sinica, 28 (1): 29-36 (in Chinese). http://www.dzxb.org/Magazine/Show?id=26517
王岩, 邵媛媛, 曹凤娟, 郭晓燕. 2016.基于因子分析的地震综合指数在辽宁地区地震预测中的应用[J].防灾减灾学报, 32 (1): 1-7. http://www.cnki.com.cn/Article/CJFDTOTAL-DDYJ201601001.htm Wang Y, Shao Y Y, Cao F J, Guo X Y. 2016. Application of comprehensive index to earthquake prediction of Liaoning using factor analysis method[J]. Journal of Disaster Prevention and Reduction, 32 (1): 1-7 (in Chinese). http://www.cnki.com.cn/Article/CJFDTOTAL-DDYJ201601001.htm
中国地震局监测预报司. 2002.强地震中短期预报方法及其效能研究[M].北京:地震出版社: 45-87. Department of Monitoring and Prediction, China Earthquake Administration. 2002. Study on Medium-and Short-Term Forecast Methods and Their Efficiency[M]. Beijing: Seismological Press: 45-87 (in Chinese).
中国地震局监测预报司. 2007.中国大陆地震序列研究[M].北京:地震出版社: 20-24. Department of Monitoring and Prediction, China Earthquake Administration. 2007. Researches on Earthquake Sequences of Chinese Mainland[M]. Beijing: Seismological Press: 20-24 (in Chinese).
Bartlett M S. 1937. Properties of sufficiency and statistical tests[J]. Proc R Soc A, 160 (901): 268-282. doi: 10.1098/rspa.1937.0109
Gunopulos D, Das G. 2001. Time series similarity measures and time series indexing[J]. Acm Sigmod Record, 31 (2): 624. https://www.researchgate.net/publication/221213566_Time_Series_Similarity_Measures_and_Time_Series_Indexing
Hill B D. 2011. The Sequential Kaiser-Meyer-Olkin Procedure as an Alternative for Determining the Number of Factors in Common-Factor Analysis: A Monte Carlo Simulation[D]. Stillwater, Oklahoma State: Oklahoma State University: 40-43. https://www.coursehero.com/file/16648753/online-learning-2016-olc-conference-special-issuepdf/
Hotelling H. 1933. Analysis of a complex of statistical variables into principal components[J]. J Educ Psychol, 24 (6): 417-441. doi: 10.1037/h0071325
Kaiser H F. 1960. The application of electronic computers to factor analysis[J]. Educ Psychol Measur, 20 (1): 141-151. doi: 10.1177/001316446002000116
Moghaddam B. 2002. Principal manifolds and probabilistic subspaces for visual recognition[J]. IEEE Trans Pattern Anal Mach Intell, 24 (6): 780-788. doi: 10.1109/TPAMI.2002.1008384