The School of Computer Science is pleased to present…
Date: Friday, May 9th, 2025
Time: 10:00 am
Location: Essex Hall, Room 122
Social Network Analysis provides a powerful framework for understanding relationships within networks by uncovering structural patterns in real-world data. This dissertation focuses on its application in recommender systems, which have evolved from basic content-based and collaborative filtering methods to more advanced graph-based techniques. Graph Neural Networks (GNNs) have significantly improved the accuracy of recommendations and personalization. Our research investigates the role of sparsity and anomalies in these systems.
The study begins by reviewing the evolution of recommender systems, with a focus on recent trends including GNN-based methods and their intersection with broader societal dynamics. We then apply outlier detection and social network analysis techniques to model and address a real-world challenge in palliative care networks. Next, we investigate how data anomalies impact different stages of session-based recommender systems. Finally, we discuss the sparsity issue in cross-domain systems and propose a cross-domain recommendation approach to address it. Through semantic alignment and clustering, we enable knowledge transfer between data-rich and data-sparse domains. The study concludes with our experimental results and the key insights drawn from the research.
Recommender Systems, Sparsity, Anomaly, Graph-based Systems
Internal Reader: Dr. Saeed Samet
Internal Reader: Dr. Dan Wu
External Reader: Dr. Mitra Mirhassani
External Examiner: Dr. Kumaraswamy Ponnambalam
Advisor(s): Dr. Ziad Kobti and Dr. Pooya Moradian Zadeh
Chair: Dr. Arezoo Emadi
