Scholarlink

<p>Debris flow prediction remains a critical yet challenging task due to limitations in accuracy, generalization, and the ability to model the complex nonlinear behavior inherent to debris flow velocity. In response to these challenges, this study presents a novel predictive framework that integrates Convolutional Neural Networks (CNN) and Gated Recurrent Units (GRU) within a deep learning architecture. To further enhance performance, the model is optimized using the Coronavirus Herd Immunity Optimizer (CHIO), a state-of-the-art metaheuristic algorithm designed to fine-tune hyperparameters, thereby improving both predictive accuracy and generalizability. The proposed CHIO-CNN-GRU model was trained and validated using 147 datasets of debris flow dynamics and systematically benchmarked against existing methods. Experimental findings highlight its excellent performance, delivering high predictive accuracy (R<sup>2</sup> = 0.9721, MAPE = 0.0944) and robustness by efficiently capturing the nonlinear physical traits of debris flow velocity. To evaluate its real-world applicability, the model was further validated with 20 additional datasets derived from flume simulation experiments. The results confirmed its strong generalization capability and practical robustness, with improved predictive accuracy (R<sup>2</sup> = 0.9225, MAPE = 0.0545) compared to existing models.This research highlights the efficacy of combining advanced deep learning architectures with intelligent optimization algorithms in solving complex geophysical prediction tasks. The CHIO-CNN-GRU model establishes a novel standard for predicting debris flow velocity, offering a reliable and adaptable tool for real-time disaster prediction and mitigation. The findings provide a robust foundation for further advancements in debris flow research and engineering applications.</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

A new CNN-GRU deep learning framework optimized by CHIO for precise prediction of debris flow velocity

  • 期刊论文

  • 发表日期：

    2025年4月11日

查看全文
  • Liangxu Shen,
  • Tianxing Ma,
  • Hao Chen,
  • Bingfeng Ye,
  • Rui Luo,
  • Hongyue Sun

摘要

Debris flow prediction remains a critical yet challenging task due to limitations in accuracy, generalization, and the ability to model the complex nonlinear behavior inherent to debris flow velocity. In response to these challenges, this study presents a novel predictive framework that integrates Convolutional Neural Networks (CNN) and Gated Recurrent Units (GRU) within a deep learning architecture. To further enhance performance, the model is optimized using the Coronavirus Herd Immunity Optimizer (CHIO), a state-of-the-art metaheuristic algorithm designed to fine-tune hyperparameters, thereby improving both predictive accuracy and generalizability. The proposed CHIO-CNN-GRU model was trained and validated using 147 datasets of debris flow dynamics and systematically benchmarked against existing methods. Experimental findings highlight its excellent performance, delivering high predictive accuracy (R2 = 0.9721, MAPE = 0.0944) and robustness by efficiently capturing the nonlinear physical traits of debris flow velocity. To evaluate its real-world applicability, the model was further validated with 20 additional datasets derived from flume simulation experiments. The results confirmed its strong generalization capability and practical robustness, with improved predictive accuracy (R2 = 0.9225, MAPE = 0.0545) compared to existing models.This research highlights the efficacy of combining advanced deep learning architectures with intelligent optimization algorithms in solving complex geophysical prediction tasks. The CHIO-CNN-GRU model establishes a novel standard for predicting debris flow velocity, offering a reliable and adaptable tool for real-time disaster prediction and mitigation. The findings provide a robust foundation for further advancements in debris flow research and engineering applications.