Desenvolvimento de um protótipo para uso de dados não estruturados em sistema de medição de desempenho

• Ano de defesa: 2022
• Autor(a) principal: Assandre, Junior Aparecido
• Orientador(a): Oprime, Pedro Carlos
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de São Carlos; Câmpus São Carlos
• Programa de Pós-Graduação: Programa de Pós-Graduação em Engenharia de Produção - PPGEP
• Palavras-chave em Português: Sistema de Medição de Desempenho; Dados Não Estruturados
• Palavras-chave em Inglês: Big Data; Big Data Analytics; Deep Learning
• Área do conhecimento CNPq: ENGENHARIAS::ENGENHARIA DE PRODUCAO::GERENCIA DE PRODUCAO
• Link de acesso: https://repositorio.ufscar.br/handle/20.500.14289/16550

Given the potential that big data can bring to organizations with the use of the large amount of unstructured data generated quickly in different formats and sources, and the scarcity of empirical research to validate its impacts and opportunities for performance measurement systems (SMD). The main objective of this thesis was to develop an SMD prototype using deep learning and big data analytics (BDA) techniques for collecting, storing and analyzing unstructured data. Among the different formats of unstructured data, video images were used, due to their great availability. As a way to present the cognitive process of the design and construction of the SMD prototype in a functional state and thus create generalizable knowledge for this class of problem, the design science research (DSR) research method was used. During the prototype development process, it was possible to identify convergences between the MDS development phases and the phases to be followed to put a big data project into operation. The analysis of these convergences together with the learning acquired in the prototype development process, highlighted the most relevant points in this type of project. The research results indicated the importance of a support infrastructure for the use of unstructured data by the SMD and the even greater dependence on the area of ​​information technology and computer science. The results also showed that, by providing the creation of performance measures and advanced analysis of unstructured data, previously unfeasible, BDA, deep learning and computer vision techniques positively influence the validity, reliability and timeliness of this resource, as well as its level. aggregation and costs. Thus, the main contribution of the thesis was to generate relevant, practical and prescriptive knowledge for the use of unstructured data from videos in SMDs, in order to contribute to their improvement and adaptation.