Scholarlink

The implementation of 5G networks has transformed communication by introducing network slicing, enabling customized performance for applications such as Enhanced Mobile Broadband (eMBB), Ultra-Reliable Low-Latency Communication (URLLC), and Massive Machine-Type Communication (mMTC). Proposed research paper presents HybridNetSlice, a deep learning framework that integrates Convolutional Neural Networks (CNNs), Long Short-Term Memory (LSTM) networks, and Attention Mechanisms to achieve 96% accuracy in classifying network slice types and their sub-slices. The model processes one-dimensional sequences of network parameters such as bandwidth, latency, and throughput capturing both spatial and temporal patterns while prioritizing critical features using an Attention Mechanism. Utilizing a custom 5G dataset of 300,000 samples, covering three primary slice types and 15 sub-slices, the framework simulates diverse use cases. Experimental results highlight the robustness of HybridNetSlice in dynamic network management, offering insights into parameter ranges for various applications. By supporting real-world scenarios such as autonomous vehicles, remote healthcare, and smart grid optimization, the model demonstrates its practical value in enhancing resource allocation and driving intelligent 5G network slicing. These advancements position HybridNetSlice as a scalable solution for real-world deployment in next-generation networks. Research extensions include real-time data integration, dataset diversification, and advanced techniques like reinforcement learning, making HybridNetSlice a scalable solution for next-generation networks.

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

HybridNetSlice: Leveraging Deep Learning for 5G Network Slice Prediction

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Ashish Lodaya,
  • Aditi Ponkshe,
  • Anjana Bharamnaikar,
  • Narasimha Shastry,
  • Sneha Varur,
  • M. Vijayalakshmi

摘要

The implementation of 5G networks has transformed communication by introducing network slicing, enabling customized performance for applications such as Enhanced Mobile Broadband (eMBB), Ultra-Reliable Low-Latency Communication (URLLC), and Massive Machine-Type Communication (mMTC). Proposed research paper presents HybridNetSlice, a deep learning framework that integrates Convolutional Neural Networks (CNNs), Long Short-Term Memory (LSTM) networks, and Attention Mechanisms to achieve 96% accuracy in classifying network slice types and their sub-slices. The model processes one-dimensional sequences of network parameters such as bandwidth, latency, and throughput capturing both spatial and temporal patterns while prioritizing critical features using an Attention Mechanism. Utilizing a custom 5G dataset of 300,000 samples, covering three primary slice types and 15 sub-slices, the framework simulates diverse use cases. Experimental results highlight the robustness of HybridNetSlice in dynamic network management, offering insights into parameter ranges for various applications. By supporting real-world scenarios such as autonomous vehicles, remote healthcare, and smart grid optimization, the model demonstrates its practical value in enhancing resource allocation and driving intelligent 5G network slicing. These advancements position HybridNetSlice as a scalable solution for real-world deployment in next-generation networks. Research extensions include real-time data integration, dataset diversification, and advanced techniques like reinforcement learning, making HybridNetSlice a scalable solution for next-generation networks.