Scholarlink

<p>Workload prediction is essential for proactive resource scheduling in cloud computing to ensure quality of service. However, existing research can not adaptively handle noisy workload data and has difficulty capturing long-range and global dependencies in non-stationary workload data. Therefore, this study proposes a novel prediction method called VNWSTG. First, VNWSTG decomposes the workload data into multiple Intrinsic Mode Functions (IMFs) by variational mode decomposition. Then, the first IMF is adaptively denoised using a fast non-dominated sorting genetic algorithm optimized wavelet transform, and the remaining IMFs are transformed by an arcsine transform. Finally, a hybrid model consisting of a temporal convolutional network with a self-attention mechanism and a gated recurrent unit is used to predict each IMF separately and fuse them into the result. Simulation results show that VNWSTG achieves higher prediction accuracy and excellent inference speed compared with several advanced prediction methods.</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

Cloud workload prediction based on wavelet transform noise reduction and a TCN-GRU hybrid model

  • 期刊论文

  • 发表日期：

    2025年9月15日

查看全文
  • Jingmeng Li,
  • Biying Zhang

摘要

Workload prediction is essential for proactive resource scheduling in cloud computing to ensure quality of service. However, existing research can not adaptively handle noisy workload data and has difficulty capturing long-range and global dependencies in non-stationary workload data. Therefore, this study proposes a novel prediction method called VNWSTG. First, VNWSTG decomposes the workload data into multiple Intrinsic Mode Functions (IMFs) by variational mode decomposition. Then, the first IMF is adaptively denoised using a fast non-dominated sorting genetic algorithm optimized wavelet transform, and the remaining IMFs are transformed by an arcsine transform. Finally, a hybrid model consisting of a temporal convolutional network with a self-attention mechanism and a gated recurrent unit is used to predict each IMF separately and fuse them into the result. Simulation results show that VNWSTG achieves higher prediction accuracy and excellent inference speed compared with several advanced prediction methods.