基于Takens理论和SVM的滑坡位移预测

针对滑坡变形时序非线性,数据量少的特点,引入Takens理论,采用支持向量机(SVM)建立其预测模型,建模过程中,比较了由不同核函数获得的SVM模型的性能,同时将SVM与RBF、El-man神经网络模型进行外推7步预测试验比较。结果表明:RBF核函数具有更好的工程实用价值;在有限样本情况下,SVM预测模型具有更好的准确性和泛化性,其7步预测平均误差率控制在5%以内,可见该方法在滑坡变形预测方面极具潜力。

Landslide Displacement Prediction Based on Takens Theory and SVM

Aimed at time series nonlinear of landslide displacement and limited quantity of samples,using Takens theory,a model for predicting landslide displacement based on support vector machines(SVM) was presented.Comparison of different SVM models with kernel functions was made based on their predicting abilities.In order to evaluate SVM,RBF network and Elman neural network models were adopted as well to predict the last seven steps landslide displacement.The results show that RBF kernel function is of better practical value;the SVM is of greater generalization and accuracy and the average relative errors are controlled in 5%,therefore,the application of SVM on landslide displacement prediction will have a good prospect.