1. Displacement Time Series Prediction Model of Landslide Based on Phase Space Reconstruction
- Author
-
Ting Yao Jiang and Shan Shan Wang
- Subjects
business.industry ,General Engineering ,Univariate ,Recurrence period density entropy ,Landslide ,Machine learning ,computer.software_genre ,Support vector regression model ,Support vector machine ,Phase space ,Entropy (information theory) ,Artificial intelligence ,Time series ,business ,computer ,Algorithm ,Mathematics - Abstract
In order to fully reveal information about landslide displacement, it was necessary to extend a time series to a higher-dimensional state space for the characteristic of univariate time series. However, in order to control the expansion of noise, an appropriate embedded dimension of phase space reconstruction was not the bigger the better. In this paper, based on the displacement time series of landslide, the phase space theory was used to build displacement time series matrix and the entropy theory was used to get the entropy. The embedded dimension of phase space reconstruction could be adjusted according to the change of entropy and feedback of displacement prediction error and a support vector regression model was created via the support vector machine’s learning. The application on Baijiabao landslide indicates that the proposed method achieves a high accuracy and stability of prediction.
- Published
- 2014