Name
Abstract | ||
|---|---|---|
Time series prediction theory and methods can be applied to many practical problems, such as the early warning of landslide hazard. Most already existing time series prediction methods cannot be effectively applied on landslide displacement prediction tasks, mainly for two problems. Firstly, the underlying dynamics of landslides cannot be properly modeled; secondly, it is difficult to perform effective long term predictions. Considering these problems, a dynamic predictor is proposed in our paper. The predictor is established on a recurrent network structure and trained by a newly proposed learning algorithm, namely echo state network. Furthermore, multi-step predictors are built based on echo state network, following different predicting strategies. Experimental results show that, the dynamic predictors perform better than static predictors, and can produce reliable multi-step ahead predictions of landslide displacements. |
Year | DOI | Venue |
|---|---|---|
2014 | 10.1007/978-3-319-12637-1_48 | Lecture Notes in Computer Science |
Keywords | DocType | Volume |
Time series prediction,landslide,recurrent neural network,echo state network,multi-step prediction | Conference | 8834 |
ISSN | Citations | PageRank |
0302-9743 | 0 | 0.34 |
References | Authors | |
5 | 5 | |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Wei Yao | 1 | 121 | 7.24 |
| Zhigang Zeng | 2 | 3962 | 234.23 |
| Cheng Lian | 3 | 36 | 5.57 |
| Huiming Tang | 4 | 57 | 13.06 |
| Tingwen Huang | 5 | 5684 | 310.24 |