Network Management Assistance through Large Language Models (LLMs)
Loading...
Date
2025-06-15
Journal Title
Journal ISSN
Volume Title
Publisher
Mohamed Boudiaf University of M'sila
Abstract
This thesis explores the transformative role of Large Language Models (LLMs) in the
domain of network optimization, particularly within the context of 5G communication
technologies. It starts by studying deep learning fundamentals and neural network ar chitectures, emphasizing the evolution and impact of Models such as GPT, BERT, and
DeepSeek. The study then examines the integration of these models into modern net working workflows, focusing on their applications in network security, task classification
as well as answering telecom-domain questions.
A core challenge addressed in this work lies in the sheer volume and complexity of
technical data in the telecommunications industry—particularly across 3GPP (3rd Gen eration Partnership Project) standards, which has overseen the development of universal
standards for Mobile Wireless Networks (MWNs). 3GPP continuously publishes a large
number of intricate documents, making it difficult for engineers and researchers to stay
updated and extract relevant information efficiently. This creates a need for advanced
methods to process, analyze, and understand these documents to ensure network relia bility and performance.
To address this, the thesis aiming to be a roadmap for researchers and practitioners
to leverage LLMs in solving various telecom tasks. Special attention is given to the appli cation of fine-tuning and Retrieval-Augmented Generation (RAG) techniques to improve
technical comprehension and automate knowledge extraction from3GPP specifications.
A practical evaluation is conducted using the TSpec-LLM dataset, and chatbot are de veloped to classify telecom tasks and respond to domain-specific queries.
Finally, The research demonstrates how LLMs will become increasingly important
for telecom-specific operations by enhancing information retrieval accuracy and efficiency
and accessibility in complex technical domains.
Description
Keywords
Chatbot, 3GPP Documents, Large Language Models (LLMs), Network Optimization, Retrieval-Augmented Generation (RAG), Fine-tuning