使用M_f值、C值、D值异常资料进行地震短期预测的神经网络方法  

The Forecasting Earthquake Method of Neural Networks in the Short term Using Anomalous Data of M f,C and D Values

作　　者：王炜[1] 戴维乐[1] 

机构地区：[1]安徽省地震局

出　　处：《中国地震》1997年第4期394-401,共8页Earthquake Research in China

摘　　要：介绍了神经网络的一些基本概念、ＢＰ神经网格及其算法。使用地震强度因子Ｍｆ值、地震空间集中度Ｃ值、地震危险度Ｄ值对华北地区１９７２～１９９２年期间进行空间扫描的中期和短期异常资料，通过ＢＰ神经网络进行学习并进行地震短期预测。研究结果表明：利用这３类资料的多项因子进行短期预测的效果较为理想。文章还对使用ＢＰ神经网络的一些具体问题进行了讨论。Some basic concepts of the neural networks, BP neural networks and its algorithm are introduced in the paper. The BP neural networks are applied to learnning and short term prediction by using of the anomaly data of seismic magnitude factor M f value, concentrative digree C value and dangerous digree D value in mid term and short term during 1972～1992 in North China. The testing results show that the short term forecasting effect is very good in using some features of the three parameters.

关 键 词：地震预报 地震活动性 专家系统 神经网络 

分 类 号：P315.7[天文地球—地震学]