神经网络模型在地震预报中的某些应用

本文介绍了人工神经网络模型以地震活动性指标为基础应用于地震预报的一些最新研究结果，选用多层前向神经网络模型及ＢＰ算法，其输入取不同的地震活动性指标的集合，输出为某一指定地区在未来时段内可能发生的最大地震的震级，以华北及首都圈地区为例，用多组不同类型的地震活动性指标进行学习与检验，结果表明，利用人工神经网络模型对未来时段震级预报的符合率较高，内检预报符合率可达１００％，外推预报符合率达到６０％以上。

Some Application of Neural Network Model to Earthquake Prediction

Some new results on the application of artificial neural networks to earthquake prediction based on seismicity indices have been described. The multi-layer forward-type modelof neural networks (NN) and the well-known BP algorithm were chosen and applied. Theset of different indices of seismic activity was used as the input of NN, The magnitude oflargest earthquake which may occur in the forthcoming time interval in a given region wastaken as the output of NN. The North China and Capital Circle regions have been taken asexamples. By using a series of different sets of predicting indices of seismicity of differenttypes to learning and checking, the results obtained show that the consilience degree ofmagnitude prediction for the forthcoming time interval by applying the method of neuralnetworks is more high, namely, the consilience degree of interior check may reach 100%,and the consilience degree of extrapolational prediction may reach above 60%.