|Thesis abstract: |
Many models have been developed in the last years that try to learn users¿ interests by fitting behavioral data with stationary models. However, few models take into account the contextual nature of the user model. For instance, user preferences drift with time, either because of the emergence of new products or because of the gradual change of the taste of the user. The aim of this research is to develop models that capture the contextual nature of tastes over time.