Big data analytics applied to sensor data of engeneering structures: predictive methods

• Autor(a) principal: Caçador, Filipe Galvão Chambel
• Data de Publicação: 2017
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10071/15483
• topic: Predictive analytics; Machine learning; Data mining; Big data; Statistical analysis; Dam monitoring; Redes neuronais; Modelos estatísticos; Métodos de previsão

Predictive models are fundamental instruments for providing dam safety analysis. They are important tools to retrieve conclusions about the structural safety of these dams. The data for these predictive models is gathered through sensors embedded within these structures. Even though predictive models are powerful tools for analysis and prediction, other machine learning and statistical models, like neural networks, have been developed over the years. Due to the many ways dam safety analyses is performed, the focus is to improve the existing methods by comparing them with each other. This work is focused on developing the methodology that compares different predictive models, like the Multiple Linear Regression Model, the Ridge Regression Model, the Principal Component Regression Model and Neural Networks, as well as comparing different re-sampling techniques for separating the data. This methodology is applied to a case study, with the purpose of finding which combinations of input variables provide the highest accuracy for predicting the behavior of these structures.