Scholarlink

<p>To create dependable quality of service (QoS) communication systems for B5G/6G that users can rely on, models must be thoroughly assessed using machine learning (ML) and deep learning (DL) techniques. This work integrates a behavior-based trust ranking model into a layer-wise wireless network design to provide trustworthy QoS communication. Every user is assigned to one of three trust rank where rank 1 is least supportive and rank 3 is most supportive for QoS communication. This work also discusses how an RS dataset is generated from behavior-based attributes and how to prepare and optimize the dataset for model training. Performance of eight models–four deep learning models (long Short-term memory (LSTM), one- and two-layer and gated recurrent unit (GRU), one- and two-layer) and four machine learning algorithms (decision tree (DT), K-nearest neighbours (KNN), random forest (RF), and naïve bayes (NB)-is assessed using accuracy, precision, recall, and F1-score across full and attribute-selected datasets. Attributes are chosen from the entire dataset to form selected attributes dataset using p-value analysis and a correlation matrix. The results highlight the significance of accountability and interpretability in AI models to guarantee trustworthy decision-making in QoS-critical settings.</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

Behavior-Driven Trust Ranking for QoS in B5G/6G Networks: A Comparative Analysis of Machine Learning and Deep Learning Models

  • 期刊论文

  • 发表日期：

    2026年6月26日

查看全文
  • Santosh Sharma,
  • Rakesh Kumar Jha,
  • Anil Kumar

摘要

To create dependable quality of service (QoS) communication systems for B5G/6G that users can rely on, models must be thoroughly assessed using machine learning (ML) and deep learning (DL) techniques. This work integrates a behavior-based trust ranking model into a layer-wise wireless network design to provide trustworthy QoS communication. Every user is assigned to one of three trust rank where rank 1 is least supportive and rank 3 is most supportive for QoS communication. This work also discusses how an RS dataset is generated from behavior-based attributes and how to prepare and optimize the dataset for model training. Performance of eight models–four deep learning models (long Short-term memory (LSTM), one- and two-layer and gated recurrent unit (GRU), one- and two-layer) and four machine learning algorithms (decision tree (DT), K-nearest neighbours (KNN), random forest (RF), and naïve bayes (NB)-is assessed using accuracy, precision, recall, and F1-score across full and attribute-selected datasets. Attributes are chosen from the entire dataset to form selected attributes dataset using p-value analysis and a correlation matrix. The results highlight the significance of accountability and interpretability in AI models to guarantee trustworthy decision-making in QoS-critical settings.