Negociación aplicada a Recomendación a Grupos: Un Estudio sobre Usuarios en el Dominio de Películas

JORGE ANDRES DIAZ PACE; SILVIA SCHIAFFINO; CHRISTIAN VILLAVICENCIO; ARIEL MONTESERIN

Cordoba

Simposio; ASAI 2017 - Simposio Argentino de Inteligencia Artificial; 2017

SADIO

Providing recommendations to groups of users hasbecome popular in many applications today. Even though thereare several group recommendation techniques, the generationof recommendations that satisfy the group members in an evenway remains a challenge. Because of this, we have developed amulti-agent approach called MAGReS that relies on negotiationtechniques to improve group recommendations. Our approachwas tested (on the movies domain) using synthetic data withsatisfactory results. Given that the results when using syntheticdata may sometimes differ with reality, we decided to assessMAGReS using data from real users. The results obtained showedfirstly that, in comparison with the recommendations producedby a traditional approach, the recommendations of MAGReSproduce a greater level of satisfaction to the group, and secondlythat the proposed approach was able to predict more accuratelythe satisfaction levels of the group members.