Filtragem Híbrida para Sistema de Recomendação de Livros utilizando Redes Neurais

This thesis addresses the development of a hybrid recommendation system for books, combining collaborative and content-based approaches, powered by Neural Networks. The increasing complexity and diversity of data make adopting hybrid models essential, and this work is justified by the need to provide readers with more relevant and personalized suggestions, overcoming the limitations of traditional models. The study aims to explore, implement, and evaluate a hybrid recommendation system using a Deep Neural Network (DNN), which not only suggests works based on past preferences but also takes into account specific literary characteristics, providing more contextual recommendations. Validation was conducted through experiments using the BookCrossing and Amazon datasets, with the evaluation comparing the accuracy of recommendations with the collaborative model and identifying improvements in the personalization of the proposed hybrid model. The main contributions of this work include implementing an effective hybrid system, enabling comparative evaluation with traditional models, and analysis of the improvements achieved.