Scholarlink

This study presents the development of a strong Artificial Neural Network (ANN) and Synthetic Minority Over-sampling Technique (SMOTE) approach to categorize Internet of Things (IoT) devices. Our strategy for addressing the issue of imbalanced datasets involves using comprehensive preprocessing techniques to normalize and standardize the data, as well as utilizing SMOTE to ensure equitable representation of each class. The artificial neural network (ANN) design is characterized by a sigmoid output layer and many dense layers with rectified linear unit (ReLU) activation. This design is specifically intended for binary classification issues. We provide results showing a classification accuracy of 92%, a mean squared error of 0.05, and an F1-score of 0.91. These results demonstrate significant improvements compared to traditional classification approaches. The proposed methodology has been demonstrated to be efficacious in a real-world Internet of Things setting by enhancing precision and reducing prejudice towards dominant categories.

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

IoT Device Classification Using Machine Learning and SMOTE

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Almuntadher Alwhelat,
  • Rahib H. Abiyev,
  • Lina Ibrahim

摘要

This study presents the development of a strong Artificial Neural Network (ANN) and Synthetic Minority Over-sampling Technique (SMOTE) approach to categorize Internet of Things (IoT) devices. Our strategy for addressing the issue of imbalanced datasets involves using comprehensive preprocessing techniques to normalize and standardize the data, as well as utilizing SMOTE to ensure equitable representation of each class. The artificial neural network (ANN) design is characterized by a sigmoid output layer and many dense layers with rectified linear unit (ReLU) activation. This design is specifically intended for binary classification issues. We provide results showing a classification accuracy of 92%, a mean squared error of 0.05, and an F1-score of 0.91. These results demonstrate significant improvements compared to traditional classification approaches. The proposed methodology has been demonstrated to be efficacious in a real-world Internet of Things setting by enhancing precision and reducing prejudice towards dominant categories.