Scholarlink

<p>The adaptive identification of the current evolution state of landslides through the analysis of landslide displacement time series data using advanced machine learning algorithms is of significant research importance. This paper proposes an advanced landslide displacement evolution state classification model based on clustering, transfer learning, and deep neural networks. The first step involves clustering analysis of the variation of landslide displacement, where landslide displacement subsequences obtained through slicing are labeled with evolution state category labels.The second step involves pre-training the deep learning model on multiple landslide displacement datasets to capture the inherent unified representations of different landslide displacement data. The third step involves fine-tuning the deep learning model on a specific landslide displacement dataset to capture the intrinsic features of the specific landslide displacement data. We validate the effectiveness of the proposed method on six landslide datasets located in China. The experimental results show that the proposed method can improve the accuracy of landslide evolution state classification on four popular deep learning models.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Adaptive classification of landslide displacement evolution states through multi-landslide data transfer learning

  • 期刊论文

  • 发表日期：

    2025年5月7日

查看全文
  • Xuhuang Du,
  • Cheng Lian,
  • Youping Li,
  • Zhiyong Qi,
  • Zhengyang Tang,
  • Jin Yuan,
  • Bo Xu,
  • Hui Zeng

摘要

The adaptive identification of the current evolution state of landslides through the analysis of landslide displacement time series data using advanced machine learning algorithms is of significant research importance. This paper proposes an advanced landslide displacement evolution state classification model based on clustering, transfer learning, and deep neural networks. The first step involves clustering analysis of the variation of landslide displacement, where landslide displacement subsequences obtained through slicing are labeled with evolution state category labels.The second step involves pre-training the deep learning model on multiple landslide displacement datasets to capture the inherent unified representations of different landslide displacement data. The third step involves fine-tuning the deep learning model on a specific landslide displacement dataset to capture the intrinsic features of the specific landslide displacement data. We validate the effectiveness of the proposed method on six landslide datasets located in China. The experimental results show that the proposed method can improve the accuracy of landslide evolution state classification on four popular deep learning models.