Open AccessResearch on the application of improved SVM model in dam displacement and deformation prediction
Author(s) -
Cheng-hui Wan,
Jiancheng Xie,
Peng Xiong
Publication year - 2020
Publication title -
iop conference series. earth and environmental science
Language(s) - English
Resource type - Journals
eISSN - 1755-1307
pISSN - 1755-1315
DOI - 10.1088/1755-1315/585/1/012047
Subject(s) - support vector machine , deformation (meteorology) , hyperparameter optimization , kernel (algebra) , displacement (psychology) , series (stratigraphy) , principal component analysis , deformation monitoring , computer science , kernel principal component analysis , radius , nonlinear system , function (biology) , mean squared error , algorithm , mathematics , kernel method , artificial intelligence , statistics , geology , psychology , paleontology , oceanography , physics , computer security , combinatorics , quantum mechanics , evolutionary biology , psychotherapist , biology
Aiming at the influence of dam deformation on many factors, the kernel principal component analysis method is used to compare the factors affecting dam deformation, and the main factors affecting dam deformation are determined. The kernel function of support vector machine is selected, the main influence factor is input, the dam deformation is output, and the optimal parameters are obtained by grid search and cross-verification to establish the model. The deformation results are studied and analyzed, and the prediction effect is evaluated by absolute error and RMSE, and compared with the standard SVM model and the Prophet method based on time series. The results show that the accuracy of support vector machine model is improved obviously by optimizing parameters such as radius of kernel function and penalty factor through mesh optimization, and the prediction accuracy of nonlinear small sample data is significantly higher than that of time series method.