Scholarlink

<p>Link quality estimation (LQE) is critical for optimizing communication in Internet of Things (IoT) networks. While previous research has explored machine learning-based LQE using Received Signal Strength Indicator (RSSI) and Packet Delivery Ratio (PDR), the potential of Decision Tree (DT) classifiers with fine-tuned hyper-parameters remains unexplored. This paper introduces LQE-DT (Link Quality Estimation– Decision Tree), a novel DT-based model that leverages advanced hyper-parameter tuning such as tree depth, minimum samples per split, and impurity criteria, ensuring optimal performance to enhance classification accuracy, computational efficiency, and predictive reliability. This paper presents a novel approach integrating RSSI and PDR with Decision Tree (DT) classifier for enhanced link quality prediction. Our proposal of LQE-DT (Link Quality Estimation- Decision Tree) offers low computational complexity and advances link quality assessment from estimation to forecasting. By proactively predicting link failures and route changes before packet loss occurs, LQE-DT improves network Quality of Experience (QoE). Our experimental evaluation using real IoT network dataset reveals that DT-based LQE achieves high precision, recall, F1 score, and accuracy with an exceptionally low training time of 0.031&#xa0;s compared to other machine learning techniques. While DTs are computationally efficient relative to Random Forests and other machine and deep learning models, our study provides new insights into the benefit of joint RSSI-PDR optimization for link quality forecasting. The results highlight the potential of LQE-DT for real time, resource-constrained IoT 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

LQE-DT: a machine learning approach to proactive link quality prediction in IoT networks

  • 期刊论文

  • 发表日期：

    2025年8月30日

查看全文
  • Safia Gul,
  • Bilal Ahmad Malik,
  • M. Tariq Banday,
  • Zahrah Ayub

摘要

Link quality estimation (LQE) is critical for optimizing communication in Internet of Things (IoT) networks. While previous research has explored machine learning-based LQE using Received Signal Strength Indicator (RSSI) and Packet Delivery Ratio (PDR), the potential of Decision Tree (DT) classifiers with fine-tuned hyper-parameters remains unexplored. This paper introduces LQE-DT (Link Quality Estimation– Decision Tree), a novel DT-based model that leverages advanced hyper-parameter tuning such as tree depth, minimum samples per split, and impurity criteria, ensuring optimal performance to enhance classification accuracy, computational efficiency, and predictive reliability. This paper presents a novel approach integrating RSSI and PDR with Decision Tree (DT) classifier for enhanced link quality prediction. Our proposal of LQE-DT (Link Quality Estimation- Decision Tree) offers low computational complexity and advances link quality assessment from estimation to forecasting. By proactively predicting link failures and route changes before packet loss occurs, LQE-DT improves network Quality of Experience (QoE). Our experimental evaluation using real IoT network dataset reveals that DT-based LQE achieves high precision, recall, F1 score, and accuracy with an exceptionally low training time of 0.031 s compared to other machine learning techniques. While DTs are computationally efficient relative to Random Forests and other machine and deep learning models, our study provides new insights into the benefit of joint RSSI-PDR optimization for link quality forecasting. The results highlight the potential of LQE-DT for real time, resource-constrained IoT applications.