Machine learning-based QoS prediction under information delay in 5G/6G networks / Ali Benhorma

Public ISBD Titre : Machine learning-based QoS prediction under information delay in 5G/6G networks Type de document : document multimédia Auteurs : Ali Benhorma, Auteur ; Zakaria Zaki, Auteur ; Radhouane Ait Mechedal, Auteur ; Lakhdar Kamel Oulad Djedid, Directeur de thèse Editeur : Laghouat : Université Amar Telidji - Département d'informatique Année de publication : 2026 Importance : 105 p. Accompagnement : 1 disque optique numérique (CD-ROM) Note générale : Option : Networks, systems and distributed applications Langues : Anglais (eng) Mots-clés : Keywords— 5G  6G  QoS Prediction  Machine learning  Information delay  CQI  Throughput  Jitter  LSTM  Transformer  XGBoost Résumé : This thesis addresses the critical challenge of predicting Quality of Service (QoS) metrics— specifically throughput and jitter—under information delay conditions in 5G and beyond networks. The Channel Quality Indicator (CQI), which serves as the primary input for QoS prediction models, experiences variable delays between the User Equipment (UE) and the base station (gNodeB), rendering the data stale by the time it reaches prediction models. This thesis systematically evaluates ten distinct machine learning techniques (XGBoost, Random Forest, LSTM, Transformer, CNN+LSTM, Lightweight Attention, CNN Spatiotemporal, Lightweight CNN, Hybrid LSTM+Stats, and Hybrid RF+GB+LSTM) across four delay scenarios (0, 5, 10, 15 steps). A novel cascaded prediction architecture is proposed, where CQI is predicted first and then used as input for throughput and jitter prediction. The thesis identifies the Hybrid RF+GB+LSTM as the best model for throughput prediction (R2 = 0.9506) and the Hybrid LSTM+Stats for jitter prediction (R2 = 0.8054). This thesis addresses three critical research gaps: (1) the information delay gap — no prior work systematically evaluates delay impact across ML techniques ; (2) the cascaded prediction gap — no framework simulates realistic CQI-first deployment ; (3) the comprehensive benchmark gap — no unified comparison exists across ten ML techniques under identical delay conditions. The problem of zero-jitter predictions caused by shared normalization in multi-output models is identified and solved through independent models and scalers. note de thèses : Mémoire de master en informatique Machine learning-based QoS prediction under information delay in 5G/6G networks [document multimédia] / Ali Benhorma, Auteur ; Zakaria Zaki, Auteur ; Radhouane Ait Mechedal, Auteur ; Lakhdar Kamel Oulad Djedid, Directeur de thèse . - Laghouat : Université Amar Telidji - Département d'informatique, 2026 . - 105 p. + 1 disque optique numérique (CD-ROM). Option : Networks, systems and distributed applications Langues : Anglais (eng) Mots-clés : Keywords— 5G  6G  QoS Prediction  Machine learning  Information delay  CQI  Throughput  Jitter  LSTM  Transformer  XGBoost Résumé : This thesis addresses the critical challenge of predicting Quality of Service (QoS) metrics— specifically throughput and jitter—under information delay conditions in 5G and beyond networks. The Channel Quality Indicator (CQI), which serves as the primary input for QoS prediction models, experiences variable delays between the User Equipment (UE) and the base station (gNodeB), rendering the data stale by the time it reaches prediction models. This thesis systematically evaluates ten distinct machine learning techniques (XGBoost, Random Forest, LSTM, Transformer, CNN+LSTM, Lightweight Attention, CNN Spatiotemporal, Lightweight CNN, Hybrid LSTM+Stats, and Hybrid RF+GB+LSTM) across four delay scenarios (0, 5, 10, 15 steps). A novel cascaded prediction architecture is proposed, where CQI is predicted first and then used as input for throughput and jitter prediction. The thesis identifies the Hybrid RF+GB+LSTM as the best model for throughput prediction (R2 = 0.9506) and the Hybrid LSTM+Stats for jitter prediction (R2 = 0.8054). This thesis addresses three critical research gaps: (1) the information delay gap — no prior work systematically evaluates delay impact across ML techniques ; (2) the cascaded prediction gap — no framework simulates realistic CQI-first deployment ; (3) the comprehensive benchmark gap — no unified comparison exists across ten ML techniques under identical delay conditions. The problem of zero-jitter predictions caused by shared normalization in multi-output models is identified and solved through independent models and scalers. note de thèses : Mémoire de master en informatique

---

1 (1 - 1 / 1) Par page : 25 50 100 200