Dissertation defense (December 14, 2023): Ryan Dutra de Abreu

Student: Ryan Dutra de Abreu

Title: A Study on the Integration of Collaborative Filtering Algorithms and Community Detection to Improve Recommendation Performance

Advisors: Laura Silva de Assis (advisor)  and Douglas de Oliveira Cardoso (co-advisor)

Committee: Laura Silva de Assis (Cefet/RJ), Douglas de Oliveira Cardoso (IPT), Diego Nunes Brandão (Cefet/RJ) and João Nuno Vinagre Marques da Silva (NESC TEC / Universidade do Porto)

Day/Time:  December 14, 2023 / 11 p.m.

Room: https://teams.microsoft.com/l/meetup-join/19%3aWfXqePNzHUyNHsKx_j1mGEKCA8uCBTA5fBbU9lGpm781%40thread.tacv2/1701372300703?context=%7b%22Tid%22%3a%228eeca404-a47d-4555-a2d4-0f3619041c9c%22%2c%22Oid%22%3a%2291e505cb-28d8-40f6-a547-adfd127dabca%22%7d

Abstract: Item recommendation in recommender systems is a widely used technique to assist users in discovering relevant content. Traditionally, collaborative filtering-based recommendation algorithms have been the most commonly used in the field due to their sound performance compared to other approaches. These algorithms aim to identify global patterns of similarity between users or items to generate recommendations. In this thesis, we investigate how combining the tasks of recommendation and community detection can lead to better recommendations than those obtained without considering implicit communities. To achieve this, we experimentally evaluate various combinations of community detection methods and recommendation algorithms, subjecting these different arrangements to both synthetic and real-world datasets.
The intrinsic goal of this effort was to unveil interesting patterns in the behavior of the resulting systems. The obtained results show that the inclusion of community detectors in the system can significantly improve both the effectiveness and efficiency of recommendation algorithms in some scenarios. These findings can be used to assist researchers and data science professionals in better understanding the benefits and limitations of this methodology. Lastly, the discoveries obtained here can be applied to a greater or lesser extent in various domains where personalized recommendations at the local level can be an effective approach to enhance the user experience, particularly in situations where there exists a large network structure and well-defined communities.