Scholarlink

<p>Wireless sensor networks are significant for a variety of applications, comprising object recognition, ambient monitoring, and data transmission. This work provides an extensive method focused on improving data analysis and model efficiency within WSNs by employing innovative feature extraction and feature selection approaches on the side of advanced machine learning algorithms. The process commences with the gathering of data from the WSN-DS dataset and WUSTL EHMS 2020 dataset, after that the gathered data undergoes precise pre-processing. Feature extraction is carried out utilizing a Weighted Kernel Principal Component Analysis based Singular Value Decomposition (WKPCA-SVD) model, which efficiently catches irregular patterns and modifies for data point significance via WKPCA, at the same time SVD manages dimensionality minimization. A state-of-the-art Reverse Wheel Selection boosted Artificial Rabbit Optimization algorithm is used for the selection of features. The last categorization is carried out utilizing the Stochastic Adaptive Gradient Hoffding Tree (SAGHT) algorithm, which incorporates the real-time adaptability of the Adaptive Hoffding Tree along with the Stochastic Gradient Descent’s optimization abilities. The performance of the model is further proven through performance metrics including precision (96.3% and 96%), sensitivity (96.1% and 94.4%), and&#xa0;specificity (96.78% and 94%). These results highlights the efficiency in improving data transmission security, at the same time keeping superior accuracy and effectiveness in WSN 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

Innovative feature extraction and selection for wireless sensor networks: enhancing cybersecurity and data transmission using WKPCA-SVD and RWS-ARO algorithms

  • 期刊论文

  • 发表日期：

    2025年7月27日

查看全文
  • G. Shanmugasundaram,
  • P. Iyappan,
  • R. Ramachandiran

摘要

Wireless sensor networks are significant for a variety of applications, comprising object recognition, ambient monitoring, and data transmission. This work provides an extensive method focused on improving data analysis and model efficiency within WSNs by employing innovative feature extraction and feature selection approaches on the side of advanced machine learning algorithms. The process commences with the gathering of data from the WSN-DS dataset and WUSTL EHMS 2020 dataset, after that the gathered data undergoes precise pre-processing. Feature extraction is carried out utilizing a Weighted Kernel Principal Component Analysis based Singular Value Decomposition (WKPCA-SVD) model, which efficiently catches irregular patterns and modifies for data point significance via WKPCA, at the same time SVD manages dimensionality minimization. A state-of-the-art Reverse Wheel Selection boosted Artificial Rabbit Optimization algorithm is used for the selection of features. The last categorization is carried out utilizing the Stochastic Adaptive Gradient Hoffding Tree (SAGHT) algorithm, which incorporates the real-time adaptability of the Adaptive Hoffding Tree along with the Stochastic Gradient Descent’s optimization abilities. The performance of the model is further proven through performance metrics including precision (96.3% and 96%), sensitivity (96.1% and 94.4%), and specificity (96.78% and 94%). These results highlights the efficiency in improving data transmission security, at the same time keeping superior accuracy and effectiveness in WSN applications.