| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 4.33 MB | Adobe PDF |
Authors
Advisor(s)
Abstract(s)
Atualmente, existe uma pandemia global, COVID-19, que pode provocar consequências
devastadoras para a saúde humana. Além disso, pessoas idosas têm uma probabilidade
maior de sofrerem os efeitos mais severos da doença. wearable devices podem monitorizar
parâmetros biológicos/físicos, e uma análise de previsão aos dados resultantes deve
permitir a identificação mais rápida possível de pessoas que possam estar infetadas com
o vírus.
Para criar um sistema com estas capacidades, desenvolveram-se os módulos health
data provider e data analysis para recolher os dados e fazer as previsões, com base nesses
dados, respetivamente. A plataforma FIWARE foi usada para implementar o backend do
sistema, através de módulos open source ready-to-use para gerir os dados no sistema.
Os médicos do Hospital da Terra Quente definiram os parâmetros biológicos/físicos
de interesse e, considerando os diferentes tipos de variáveis, implementaram-se três tipos
distintos de previsões: última observação, estimador linear de tendência e método aditivo
Holt-Winters. As previsões obtidas são usadas para classificar o estado de uma pessoa,
para um cenário de pior caso possível: provável que esteja infetado com o vírus ou provável
que não esteja infetado com o vírus. As regras de classificação foram definidas pelos
médicos do Hospital da Terra Quente.
Os resultados obtidos são promissores e trabalho futuro pode ser feito para melhorar
o sistema na inclusão de mais fontes de dados, ajuste dos métodos de previsão e na
introdução de novas metodologias de classificação final.
Nowadays, there exists a global pandemic, COVID-19, that may provoke devastating consequences to human health. Moreover, elder people have a higher probability to suffer the most harsh effects of the disease. By monitoring biological/physical parameters with the aid of wearable devices and applying forecast methods to the collected data, it should be feasible to identify, as soon as possible, people that may be infected with the virus. To create a system with such capabilities, a health data provider and a data analysis modules were implemented, to fetch the biological/physical data and to generate forecasts based on that data, respectively. The backend of the system was implemented through the FIWARE platform, using open source ready-to-use modules of this platform to manage the data in the system. The biological/physical parameters of interest were defined by the medical personnel of Hospital da Terra Quente and, considering the different type of variables in analysis, three types of forecasts were implemented: last observation, linear trend estimator and additive Holt-Winters method. The obtained forecasts are used, combined with a worst case scenario approach to classify the state of a person: likely to have been infected by the virus or not likely to have been infected by the virus. The classification rules were devised by the Hospital da Terra Quente medical staff. The obtained results are promising and future work can be done to improve the system in terms of providing extra data sources, fine tuning of the forecast methods and by introducing new final classification methodologies.
Nowadays, there exists a global pandemic, COVID-19, that may provoke devastating consequences to human health. Moreover, elder people have a higher probability to suffer the most harsh effects of the disease. By monitoring biological/physical parameters with the aid of wearable devices and applying forecast methods to the collected data, it should be feasible to identify, as soon as possible, people that may be infected with the virus. To create a system with such capabilities, a health data provider and a data analysis modules were implemented, to fetch the biological/physical data and to generate forecasts based on that data, respectively. The backend of the system was implemented through the FIWARE platform, using open source ready-to-use modules of this platform to manage the data in the system. The biological/physical parameters of interest were defined by the medical personnel of Hospital da Terra Quente and, considering the different type of variables in analysis, three types of forecasts were implemented: last observation, linear trend estimator and additive Holt-Winters method. The obtained forecasts are used, combined with a worst case scenario approach to classify the state of a person: likely to have been infected by the virus or not likely to have been infected by the virus. The classification rules were devised by the Hospital da Terra Quente medical staff. The obtained results are promising and future work can be done to improve the system in terms of providing extra data sources, fine tuning of the forecast methods and by introducing new final classification methodologies.
Description
Mestrado em Engenharia Industrial ESTG-IPB
Keywords
COVID-19 Fiware Dispositivos wearable Previsão