摘要
针对滑坡位移时间序列的非线性特性,引入基于相空间重构和最小二乘支持向量机(LSSVM)的预测法。利用Cao氏方法确定嵌入维数,根据互信息法计算最佳延迟时间;然后在相空间中,利用最小二乘支持向量机(LSSVM)建立预测模型,对滑坡进行了实证计算,且与LSSVM模型和BP神经网络模型进行了比较。结果表明,模型具有较高的精度,是科学可行的。
In view of the nonlinear characteristics of landslide displacement time sequence,introduced the prediction method based on phase space reconstruction and least squares support vector machine(LSSVM).Used Cao's method to determine the embedding dimension,according to mutual information method to compute the best delay time;then in the phase space,used least squares support vector machine(LSSVM) to establish the forecast model to compared with LSSVM and the neural network predicting mode.The test result shows that the model has the high precision,is scientific and feasible.
出处
《地理空间信息》
2011年第1期139-142,14,共4页
Geospatial Information
基金
东华理工大学校长基金资助项目(DHXK1010)
江西省数字国土重点实验室开放基金资助项目(DLLJ201014)
关键词
滑坡预测
相空间重构
最小二乘支持向量机
神经网络
landslide prediction,phase space reconstruction,least squares support vector machine,neural network
