Scholarlink

<p>Advances in AI have transformed the traditional Internet of Things (IoT) into smarter, more versatile next-generation devices. Enhanced connectivity among sensors, actuators, and appliances boosts data availability and resource control in Internet of Things networks. However, the increasing number of decentralized IoT devices leads to exponential growth in service requests, creating significant adaptation challenges. Machine learning has become essential in IoT applications, and hyperparameter optimization is crucial to model performance. Existing optimization methods face challenges such as high uncertainty, poor model generalization, and high computational costs. This study focuses on two key aspects of hyperparameter optimization for fog computing node performance prediction: cross-validation techniques and search space methods. In this research, a new optimization approach, Good Point Set Stepwise Shrinkage, is proposed to reduce uncertainty, enhance model generalization, and lower computational costs. Using fog computing performance data from Internet of Things devices, the proposed scheme optimized hyperparameters for Support Vector Machine, Back Propagation Network, and Convolutional Neural Network models. The test results show Mean Squared Error of 4.061, 4.114, and 3.963 ± 0.0323, respectively, for the Support Vector Machine, Back Propagation Network, and Convolutional Neural Network models. Good Point Set Stepwise Shrinkage resolves cross-validation uncertainties and improves randomness in the search space. Compared with Sequential Uniform Designs methods, Good Point Set Stepwise Shrinkage offers a simpler approach to constructing a unified search space. Good Point Set Stepwise Shrinkage becomes a highly suitable approach for hyperparameter optimization in fog computing performance prediction for the Internet of Things.</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 good point set stepwise shrinkage optimization in machine learning model for fog node performance prediction

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年3月18日

查看全文
  • Zhang Bo,
  • Mohammad Kamrul Hasan,
  • Elankovan A. Sundararajan,
  • Shayla Islam,
  • Peiying Zhang,
  • Fatima Rayan Awad Ahmed,
  • Nissrein Babiker Mohammed Babiker

摘要

Advances in AI have transformed the traditional Internet of Things (IoT) into smarter, more versatile next-generation devices. Enhanced connectivity among sensors, actuators, and appliances boosts data availability and resource control in Internet of Things networks. However, the increasing number of decentralized IoT devices leads to exponential growth in service requests, creating significant adaptation challenges. Machine learning has become essential in IoT applications, and hyperparameter optimization is crucial to model performance. Existing optimization methods face challenges such as high uncertainty, poor model generalization, and high computational costs. This study focuses on two key aspects of hyperparameter optimization for fog computing node performance prediction: cross-validation techniques and search space methods. In this research, a new optimization approach, Good Point Set Stepwise Shrinkage, is proposed to reduce uncertainty, enhance model generalization, and lower computational costs. Using fog computing performance data from Internet of Things devices, the proposed scheme optimized hyperparameters for Support Vector Machine, Back Propagation Network, and Convolutional Neural Network models. The test results show Mean Squared Error of 4.061, 4.114, and 3.963 ± 0.0323, respectively, for the Support Vector Machine, Back Propagation Network, and Convolutional Neural Network models. Good Point Set Stepwise Shrinkage resolves cross-validation uncertainties and improves randomness in the search space. Compared with Sequential Uniform Designs methods, Good Point Set Stepwise Shrinkage offers a simpler approach to constructing a unified search space. Good Point Set Stepwise Shrinkage becomes a highly suitable approach for hyperparameter optimization in fog computing performance prediction for the Internet of Things.