Preparação de dados e desenvolvimento de aplicação modular para a gestão de hipertensão e diabetes

Abstract

A presença de valores em falta e de anomalias são problemas que são comuns e que estão presentes em várias séries temporais. Estes prejudicam alguns métodos e algoritmos de análise de dados, sendo necessário fazer um pré-processamento dos dados que elimine as anomalias e que faça a substituição dos valores em falta por valores razoáveis. O foco desta dissertação é o desenvolvimento e testagem de um módulo de pré-processamento de dados que contenha algoritmos que façam a deteção de anomalias e imputação de valores em falta, em séries temporais com dados de pressão sanguínea e glicose. Foi desenvolvido um estudo, após a seleção de potencias algoritmos para resolverem os problemas expostos, que visa avaliar e posteriormente determinar os melhores métodos. Uma aplicação demonstrativa foi criada para expor o trabalho realizado em conjunto com uma interface a ser entregue para o projeto “POWER - Empowering a digital future".aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa

The presence of missing data and outliers are problems that are common and are present in various time series. This problems harm some methods and algorithms of data analysis, so it is necessary to do a pre-processing of the data, in away that eliminates outliers and substitutes the values that are missing for reasonable values. The focus of this dissertation is the development and testing of a data pre-processing module that contains algorithms that do outlier detection and missing value imputation, in time series with data about blood pressure and glucose. It was developed a study, after the selection of potencial algorithms to solve the problems already mentioned, to evaluate and posteriorly determine the best methods. An demonstrative application was created to expose the work done and it was developed an interface to be delivered to the project “POWER - Empowering a digital future”.aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa