Recomendação de carreira de pesquisadores :uma abordagem baseada em personalização, similaridade de perfil e reputação

• Autor(a) principal: Vivian, Gláucio Ricardo
• Data de Publicação: 2017
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Biblioteca de teses e dissertações da Universidade de Passo Fundo (BDTD UPF)
• Texto Completo: http://tede.upf.br/jspui/handle/tede/1425
• topic: Pesquisadores; Pesquisa; Sistemas de computação; CIENCIA DA COMPUTACAO::SISTEMAS DE COMPUTACAO

Traditional Recommender Systems seek to assist users in the selection of products and content. In acontemporaneous environment, with high offers of informations, this aid can be the differential betweensuccess or fail. In the field of scientific research, the reality of researchers are converging to significantincrease in the quantity and diversity of production. I addition of the traditional publications in papers formats, there are numerous other forms of production that are gradually being stimulated. Among many,we was possible to cite: patents, softwares, advisory, reviews, publishing, books, research projects and network of co-authorship. This paradigm tax to researchers, makes it more complex and arduous the task of researchers plan career projection. In this context, the recommender approach can supportthe researchers, seeking to guide them with effective recommendations strategies in the planning oftheir career. In others words, a recommendations approach may suggest to the researcher what, howand when to perform a particular production. As a result, one has the possibility to be performing themost appropriate activity and in the most appropriate chronological order. The objective of this workis to propose a recommendation approach to contribute to the career management of researchers, as well as support for research groups, post graduate programs and institutions, to follow the evolution ofa researcher’s scientific reputation. For that, the profile similarity and academic reputation was used asthe premise recommendation. The experiments were performed with CNPq Productivity Researchersin the areas of Computer Science, Dentistry and Economics. It was observed that the proposed approach have a good coverage in the generation of recommendations, especially for researchers with lower reputations (test groups and initial levels of CNPq). We also observed an excellent diversity in the recommended items, which indicates a low repetitions of similar recommendations (same item)