Modelagem e implementação de sistemas de recomendação de serviços públicos : personalizando o governo digital para o cidadão do estado de Mato Grosso

• Autor(a) principal: Campos, Sandro Luís Brandão
• Data de Publicação: 2021
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Repositório Institucional da UFMT
• Texto Completo: http://ri.ufmt.br/handle/1/5020
• topic: Recomendação; Personalização; Filtragem; Governo digital; Serviços públicos; CNPQ::CIENCIAS SOCIAIS APLICADAS::ADMINISTRACAO; Recommendation; Personalization; Filtering; Digital government; Public services

The application of recommendation systems technologies, currently well established in the market as a mechanism of strong relationship with the consumer, is a model of success in several business environments. However, it is little explored in digital government scenarios, especially in terms of strengthening the relationship between public administration and citizens. The central aspect of this study focused on the application of recommendation systems technologies in digital government services aimed at citizens, with the implementation of machine learning algorithms based on citizens' access to public services, and their metadata, personalizing their journey and recommending other services and information due to the similarity between the data. To meet the demands of this context, several discussions needed to be held and structured for the viability of the solution and a recommendation platform was proposed. Thus, the main contributions were the evaluation of public policy recommendation systems, management of public services portfolio, design and implementation of service recommendations interface, modeling for structuring recommendation strategies, privacy and data governance related to the model recommendations by the public administration and even the hybrid recommendation generation model itself based on content filtering. To validate the hybrid filtering proposal, tests were carried out considering collaborative filtering, contentbased filtering, and filtering by popularity, but filtering models of the most accessed services, most relevant services and news related to the Internet were still considered. citizen. A base of real interactions between citizens and services was worked out with 115897 records, considering 97261 citizens accessing 71 possible services. Although the central focus was the implementation of the recommendation model, other products were structured and implemented to support the solution's viability, thus resulting in 14 concrete contributions as a result of the entire study.