潜在地震灾害预测和评价神经网络系统及其若干问题

结合人工神经网络自身的特性和地震灾害预测研究的特点,本文应用神经网络模型,建立了潜在地震灾害预测和评价系统。针对网络模型参数设置、数据归一化、中间层神经元最优数目以及泛化分类评价指标等若干实际问题给出了实际可行的解决方案。通过大样本数据对网络的训练,形成了有识别和记忆功能的非线性预测和评价系统。对网络的测试和检验,论证了该系统在预测潜在地震灾害上的可行性和有效性。同时,从测试精度出发,探讨了这种预测网络存在的不足,并给出了相应的改进建议,为开展进一步的研究工作提供了参考。

Artificial Neural Network System for Potential Seismic Disaster Forecasting and Evaluation and Some Related Problems

Applying artificial neural network(ANN),the paper erected forecasting and evaluation system for potential seismic disaster based on the characteristics of ANN and forecasting research.Some practical measures for parameter setting,data normalization,optimum number of interface layer neurons and generalization of index classification during system modeling have been given.Through training the network using large sample data,a non-linear forecasting system with identifiable and remembrance is formed.By testing and verification,this forecasting system has been proved to be feasible and effective.The precision of forecasting,however,is not so excellent,it needs further investigation.Some suggestions are proposed in this direction.