Slope Deformation Prediction Based on IVDF-SVR Coupling Model
Given the shortcomings of improved variable-dimension fractal in the fitting and prediction of fractal dimensions,this paper proposes a fractal prediction model based on the coupling of the improved variable-dimension fractal (IVDF) theory and the support vector regression (SVR) machine theory,namel...
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | zho |
| Published: |
Editorial Office of Pearl River
2022-01-01
|
| Series: | Renmin Zhujiang |
| Subjects: | |
| Online Access: | http://www.renminzhujiang.cn/thesisDetails#10.3969/j.issn.1001-9235.2022.05.011 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Given the shortcomings of improved variable-dimension fractal in the fitting and prediction of fractal dimensions,this paper proposes a fractal prediction model based on the coupling of the improved variable-dimension fractal (IVDF) theory and the support vector regression (SVR) machine theory,namely the IVDF-SVR coupling model.The model uses the SVR machine theory to fit and predict the fractal dimension sequence in the original improved variable-dimension fractal model.This paper takes the slope displacement monitoring data of Maoping Landslide as an example and selects the dimensioned segmented ln(r)-ln(S1) curve of the cumulative sum sequence as the fractal parameter curve of the prediction model.The IVDF model is utilized to calculate the segmented fractal dimension of each curve,and slope displacement is predicted.Then,the IVDF-SVR coupling model is used for another round of calculation and prediction.The prediction results show that the IVDF-SVR coupling model makes full use of the self-similarity in the fractal theory so that the prediction model has strong noise resistance.Moreover,by incorporating the self-learning ability in the SVR theory,it enables data fitting and prediction under small samples and nonlinear conditions.These advantages provide the proposed model with a favorable prediction length,high prediction accuracy,and ultimately a bright application prospect. |
|---|---|
| ISSN: | 1001-9235 |