Scholarlink

<p>In ground subsidence analysis, the application of Variational Mode Decomposition (VMD) algorithms and Transformer models has been increasingly adopted. However, the tuning of their hyperparameters involves complexity and substantially influences predictive accuracy. To address this issue, an NRBO algorithm is proposed to perform real-time and fine-grained adjustments of hyperparameters. Furthermore, an NRBO–VMD–NRBO–Transformer hybrid model has been developed for the purpose of predicting ground subsidence. Model validation conducted in Xining City indicates that, across long-term, medium-term, and short-term time series analyses, the proposed model outperforms both the standalone Transformer model and the NRBO-optimized VMD–Transformer model. At monitoring point JC23, the R<sup>2</sup> values for the respective time series are observed to reach 0.8622, 0.9011, and 0.9268. The NRBO–VMD–NRBO–Transformer model exhibits strong predictive performance in subsidence forecasting. In addition to effectively capturing subsidence trends, the model demonstrates robustness, stability, generalization ability, strong positive correlation, and adaptive parameter tuning capabilities based on data volume.</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

Utilizing GNSS data and NRBO-VMD-NRBO-transformer for accurate settlement prediction in Nanshan, Xining

  • 期刊论文

  • 发表日期：

    2025年6月10日

查看全文
  • Xiangxiang Hu,
  • Biao Li,
  • Jixin Yin,
  • Yaya Shi,
  • Bao Song,
  • Dongdong Pang,
  • Baokang Liu

摘要

In ground subsidence analysis, the application of Variational Mode Decomposition (VMD) algorithms and Transformer models has been increasingly adopted. However, the tuning of their hyperparameters involves complexity and substantially influences predictive accuracy. To address this issue, an NRBO algorithm is proposed to perform real-time and fine-grained adjustments of hyperparameters. Furthermore, an NRBO–VMD–NRBO–Transformer hybrid model has been developed for the purpose of predicting ground subsidence. Model validation conducted in Xining City indicates that, across long-term, medium-term, and short-term time series analyses, the proposed model outperforms both the standalone Transformer model and the NRBO-optimized VMD–Transformer model. At monitoring point JC23, the R2 values for the respective time series are observed to reach 0.8622, 0.9011, and 0.9268. The NRBO–VMD–NRBO–Transformer model exhibits strong predictive performance in subsidence forecasting. In addition to effectively capturing subsidence trends, the model demonstrates robustness, stability, generalization ability, strong positive correlation, and adaptive parameter tuning capabilities based on data volume.