Regresar

Enhanced Books Recommendation Using Clustering Techniques and Knowledge Graphs

Abstract:

Nowadays, there is an increased interest in recommender systems in different fields, since they allow personalizing the delivery of content according to the needs and preferences of each user. This paper proposes a recommendation approach that suggests books according to their metadata and explicit feedback given by users to them. First, we reuse a public dataset of books and use k-means to identify different groups of books based on their information. Second, we analyze the variation between these groups to pbkp_redict the rating that a specific user would give a book. Third, we use the collaborative filtering technique to represent books, using the information of the user’s rating and the group of the book to be recommended. Finally, users will receive an enhanced explainable response, i.e. a list of books with relevant metadata, i.e., the output could be helpful for users to understand the delivered recommendations. The original book’s metadata was enriched using information available on DBPedia, a well-known RDF-based Knowledge Graph fed from Wikipedia. Results include experiments of clustering and pbkp_rediction techniques because our approach considers the information of the user’s ratings and the available information of the books to the recommendation, facilitating the understanding of the outputs.This approach achieves a good performance using Precision, Recall, and F-measure to measure the quality of the recommendations, and MAE to measure the pbkp_rediction’s accuracy.