Services on Demand
Journal
Article
Indicators
- Cited by SciELO
- Access statistics
Related links
- Similars in SciELO
Share
RISTI - Revista Ibérica de Sistemas e Tecnologias de Informação
Print version ISSN 1646-9895
Abstract
AGUIAR, Janderson Jason B.; ARAUJO, Joseana M. F. R. de and COSTA, Evandro de B.. Comparative Study of Approaches for Recommender Systems based on Personality using IBM Watson Personality Insights. RISTI [online]. 2020, n.40, pp.73-88. Epub Dec 31, 2020. ISSN 1646-9895. https://doi.org/10.17013/risti.40.73-88.
Considering the personality of users in Recommender Systems can provide more relevant results. In this study, we analyzed whether, with advances in personality detection (inference without using questionnaires), Collaborative Filtering approaches based on personality continue to improve the accuracy of traditional approaches (based essentially on ratings). Furthermore, we analyze whether there are differences when applying different personality models. We considered 1058 TripAdvisor users and 10889 Amazon customers in the experiment, with personality characteristics inferred via IBM Watson Personality Insights. The results indicated the possibility of improving accuracy by employing an approach using inferred data concerning any personality models analyzed (Big Five, Needs, and Values). Moreover, the Values model provided results equivalent to the Big Five model (without facets), and, in general terms, there was no improvement when using the Big Five model with data from its facets (nor when including data from the other models).
Keywords : Recommender Systems; Collaborative Filtering; Personality-based Recommendation; Watson Personality Insights..