Back to search

École de technologie supérieure

5G data generation for machine learning applications

Abstract

dc:description

Fifth-generation mobile networks enable unprecedented throughput, ultra-low latency, and massive device connectivity, giving rise to traffic patterns that are highly dynamic and complex. Optimizing such networks increasingly relies on machine learning, yet effective models require large-scale, complete, and representative datasets. To address this challenge, this work relies on large-scale urban mobility datasets, chosen for their traffic patterns and temporal dynamics, which closely reflect those observed in 5G networks. However, these datasets often suffer from missing observations, which motivates the development of an adaptive imputation approach designed to reconstruct incomplete time series while preserving seasonal structures and temporal dependencies. To evaluate imputation quality where ground truth is unavailable, novel metrics are introduced, providing an alternative to traditional metrics and enabling a more robust and informative evaluation. After validation, the reconstructed urban datasets are transformed into preliminary 5G traffic traces, providing a structured foundation for further modeling and analysis. Building on this foundation, a principled traffic generation framework based on the Maximum Entropy Principle is proposed, which encodes empirically observed constraints to produce realistic synthetic data that accurately capture temporal dynamics and distributional properties. This approach is particularly effective in real-world scenarios where access to highquality 5G traffic data is limited due to privacy concerns, operational constraints, or incomplete measurements. By generating representative and statistically consistent synthetic data, the framework facilitates improved forecasting, anomaly detection, and network performance evaluation, enabling telecommunication operators and researchers to design, test, and optimize next-generation mobile systems more effectively. Extensive experiments confirm that the proposed framework improves data quality, supports realistic 5G traffic modeling, and enables more robust forecasting, anomaly detection, and network performance evaluation in nextgeneration mobile systems.

Degree

thesis:*
Name thesis:degree_name
Mémoire de maîtrise électronique
Level thesis:degree_level
masters
Grantor dc:publisher
École de technologie supérieure
Year dc:date
2025

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Farjallah, Rania

Subjects

dc:subject × 1

Rights

Language dc:language
en

Identifiers

dc:identifier.*
Identifier
INSERT-YOUR-LIBRARY-ID-HERE3737

Chain of custody

source
Harvested from
ETS (Quebec)
Base URL
espace.etsmtl.ca/cgi/oai2
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Farjallah, Rania. 5G data generation for machine learning applications. masters thesis, École de technologie supérieure, 2025.