径向基神经网络在地面沉降预测中的应用

基于MATLAB6.5平台编程,利用前四年的沉降量作为输入神经元,后一年的沉降量作为输出神经元,重复此过程,构建了上海高桥地区地面沉降预测径向基神经网络.以历史沉降数据为训练样本,并对其进行归一化处理,在此基础上,采用未归一化及归一化后的训练样本进行网络训练与检验.结果表明,归一化后的训练样本训练得到的径向基网络具有良好的预测性能.最后利用该网络对1990-2010年的地面沉降量进行了预测.

Application on Land Subsidence Forecast by RBF Network

Based on MATLAB6.5 programming,RBF neural network of land subsidence forecast in the area of Shanghai Gaoqiao was constructed by the past four years' subsidence values as input neural cell and the later year value as output neural cell and repeating the process.According to former subsidence data which were taken as training data are normalized,the networks respectively using non-normalized training data and normalized training data were trained and verified,the results shown RBF network whose training data was normalized has better forecast capability,at last the land subsidence values from 1990 to 2010 were forecasted by utilize this network.