Ευφυή συστήματα συστάσεων : τεχνικές και υλοποίηση
Intelligent recommendation systems : techniques and implementation
View/
Keywords
Συστήματα συστάσεων ; Συνεργατική διήθηση ; Βαθιά μάθηση ; Γραφικά νευρωνικά δίκτυα ; Παραγοντοποίηση πίνακα ; Μηχανισμοί Προσοχής ; Δικαιοσύνη αλγορίθμωνAbstract
The rapid growth in the volume of available digital information has established Recommendation Systems as one of the most critical application domains of artificial intelligence, with use cases spanning e-commerce, streaming services, education, healthcare, and agriculture. This thesis conducts a systematic literature review of one hundred recent scientific publications, primarily from the 2020-2026 period, aiming to holistically map the evolution, architectures, and challenges of the field. The work begins with classical collaborative filtering and matrix factorization methods, progresses to deep learning architectures (such as autoencoders, Neural Collaborative Filtering, and Self-Attention mechanisms) and extensively examines Graph Neural Networks, which represent user-item interactions as a graph, enabling more expressive modeling of complex relationships. Particular emphasis is placed on specialized categories of recommender systems (context-aware, multi-behavior, explainable, fairness-aware), as well as on the main application domains, ranging from traditional e-commerce to emerging fields such as healthcare, agriculture, and cybersecurity. The thesis also critically analyzes the central challenges of the field (data sparsity, cold start, popularity bias, privacy concerns, and robustness against adversarial attacks) and presents a comparative evaluation of the main architectures based on results reported in the literature. The findings reveal a clear trend of transition from classical approaches toward deep learning and graph-based architectures, alongside growing interest in fairness, interpretability, and responsible artificial intelligence, laying the groundwork for future research directions such as the integration of Large Language Models and on-device personalization architectures.
Department
Σχολή Τεχνολογιών Πληροφορικής και Επικοινωνιών. Τμήμα ΠληροφορικήςNumber of pages
108Language
GreekCollections
Except where otherwise noted, this item's license is described as
Αναφορά Δημιουργού-Μη Εμπορική Χρήση-Όχι Παράγωγα Έργα 3.0 Ελλάδα
Αναφορά Δημιουργού-Μη Εμπορική Χρήση-Όχι Παράγωγα Έργα 3.0 Ελλάδα