MMRecommender: arquitetura aberta para sistemas de recomendação

Abstract

Recommender Systems can be defined as systems capable of recommending resources adhering to user or group of user’s profile and context, and can be applied in several domains, such as education, tourism and e-science. Due to this adaptive feature, it is possible to find several recommender models in the literature, each with combinations of different methods and algorithms. This variety of recommendation models can make it difficult to implement Recommender Systems. In this scenario, the present dissertation presents an open architecture MMRecommender, where through the combination of components present in each step it is possible to instantiate recommender models that can be applied to several application domains. To evaluate the architecture, three case studies are presented under different perspectives: the first case study focuses on the adaptation of an existing Recommender System to the MMRecommender architecture, the second case study implements a recommender model created from the proposed architecture in a scientific software ecosystem, and finally a case study evidences how the proposed architecture made possible the implementation of a Tourist Recommender System used in the RIO 2016 Olympic Games. After evaluating each case study it was possible to identify indications that the proposed architecture can help in the construction of recommender models.