User Models Development Based on Cross-Domain for Recommender Systems
Marivaldo Bispo Rodrigues, Gabriela O. Mota da Silva, Frederico Araújo Dur�ão · 2016
Recommender systems are used by many sites and services, and are important tools to help the user to find what is most relevant in the immense amount of information available. One way to build a Recommendation System is content-based filtering, which recommends items to the user based on a profile that contains information about the content, such as genre, keywords, etc. For this to happen effectively, the system must take into account the preferences and needs of users in order to generate useful recommendations. This work proposes the modeling of user profiles with integration of multiple domains and automatically. Then, through a transfer of knowledge of a domain to another, increase the performance of the recomendation. The results of the evaluation showed that information sharing between the domains increased the performance of the recommendation, as in the test with the metric [email protected], where obtained an improvement of more than 90\%.