Hyperml and deep interest network to build a recommender system for modatta: data privacy with federated learning

• Autor(a) principal: Aguiar, Emila Cristina Ribeiro
• Data de Publicação: 2021
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10362/140157
• topic: Machine learning; Deep learning; Hyperbolic embeddings; Data monetization; Recommender system; Federated learning; Data privacy; Business analysis; Domínio/Área Científica::Ciências Sociais::Economia e Gestão

The Work Project recognized two Deep Learning approaches intended to learn embedding representations, which ultimately lead to the assemble of a Recommender System to obtain additional insights from users’ profile and liking preferences. The following paper consists in the development of a prototype model aiming at demonstrating an alternate approach to training data, where several users collaboratively train a model, using a Machine Learning method, Federated Learning. It proposes to perform further analysis of its results and of the benefits of its employment in companies like Modatta, whose desire is to bring solutions to the current challenge in data privacy.