| 3 |
2025/09/26 04:32 |
屈*锋 |
Paper title：Landslide displacement prediction based on the CEEMDAN method and Transformer model. Paper abstract：Landslides are a typical geological disaster and pose a great challenge to land use management. However, traditional landslide displacement models often neglect the impact of random displacements. To address this issue, this paper proposes a deep learning model based on the Transformer architecture. The model innovatively applies the CEEMDAN method to decompose landslide displacements and employs the GRA-MIC fusion correlation calculation method to identify the influencing factors of displacement. Finally, the Transformer model is used for prediction. By leveraging the self-attention mechanism, the constructed Transformer model comprehensively captures the long-term global dependencies of displacement sequences for landslide displacement prediction. Two new concepts for evaluating landslide states and landslide trends are introduced to further enhance the predictive performance of the model. Subsequently, we discuss the predictive performance of the Transformer model under four different input conditions and compare it with seven other predictive models (including CNN-BiLSTM). The experimental results demonstrate that the proposed model has strong applicability in landslide-prone areas, providing solid support for disaster prevention, mitigation, and land use management. Keywords: land use management; landslide displacement prediction; complete ensemble empirical mode decomposition with adaptive noise; transformer Paper type：研究型论文 Tutor 任青阳
|
本作品采用
知识共享署名
4.0 国际许可协议进行许可。
CSTR:11738.11.ncdc.Sanxia.2020.89
10.12072/sanxia.020.2018.db
