A carregar...
Publicação
Impacto demográfico e financeiro da pandemia Covid 19 em Portugal : previsão do número de mortes e do PSI
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 1.94 MB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
A pandemia de COVID-19 é uma das maiores crises de saúde do século XXI,
afetou completamente o quotidiano da sociedade e impactou toda a população mundial,
económica e socialmente. O uso de algoritmos de machine learning para o estudo de
dados relativamente a esta pandemia tem sido bastante frequente nos mais variados
artigos publicados nos últimos tempos. Nesta dissertação foi analisado o impacto de
diversas variáveis (número de casos, temperatura, pessoas totalmente vacinadas, número
de vacinações diárias e vários indicadores da mobilidade) no número de mortes causadas
pela COVID-19 ou SARS-CoV-2 em Portugal e no índice da bolsa Portuguesa, o PSI, de
forma a encontrar os modelos preditivos mais adequados. Foram utilizados vários
algoritmos, como o OLS, Ridge, MLP, Gradient Boosting e Random Forest através do
software de programação Python. A análise foi dividida em dois modelos, o primeiro
referente à previsão do número de mortes causadas pela COVID-19 e o segundo à
previsão do PSI. No primeiro modelo foram usadas as variáveis originais, enquanto no
segundo modelo foi feita uma Análise de Componentes Principais, que posteriormente
foram usados para a regressão do modelo. O método utilizado para o processamento dos
dados foi o CRISP-DM. Os dados foram obtidos através de uma base de dados pública.
Por último, referir, que o Gradient Boosting foi o que obteve melhores resultados para
ambos os modelos, de acordo com as métricas de precisão utilizadas. É de salientar
também a maior eficácia dos algoritmos de Ensemble e de redes neuronais em
comparação com os algoritmos lineares na previsão dos dados utilizados.
The COVID-19 pandemic is one of the biggest health crises of the 21st century, it has completely affected society’s daily life, and has impacted populations worldwide, both economically and socially. The use of machine learning algorithms to study data from the COVID-19 pandemic has been quite frequent in the most varied articles published in recent times. In this dissertation it was analyzed the impact of several variables (number of cases, temperature, people fully vaccinated, number of daily vaccinations and several mobility variables) on the number of deaths caused by COVID19 or SARS-CoV-2 in Portugal and on the number of the Portuguese stock index, PSI, to find the most appropriate predictive model. Several algorithms were used, such as OLS, Ridge, MLP, Gradient Boosting and Random Forest through Python programming software. The analysis was divided into two models, the first referring to the prediction of the number of deaths caused by COVID-19 and the second to the PSI prediction. In the first model, the original variables were used, while in the second model, a Principal Component Analysis was made, that were later used for the regression of the model. The method used for data processing was CRISP-DM. Data were obtained from an open access database. Finally, it should be noted that Gradient Boosting was the algorithm that obtained the best results according to the precision metrics that were used. It is worth highlighting the greater efficiency of the Ensemble and neural networks algorithms compared to the linear algorithms in the prediction of the data used.
The COVID-19 pandemic is one of the biggest health crises of the 21st century, it has completely affected society’s daily life, and has impacted populations worldwide, both economically and socially. The use of machine learning algorithms to study data from the COVID-19 pandemic has been quite frequent in the most varied articles published in recent times. In this dissertation it was analyzed the impact of several variables (number of cases, temperature, people fully vaccinated, number of daily vaccinations and several mobility variables) on the number of deaths caused by COVID19 or SARS-CoV-2 in Portugal and on the number of the Portuguese stock index, PSI, to find the most appropriate predictive model. Several algorithms were used, such as OLS, Ridge, MLP, Gradient Boosting and Random Forest through Python programming software. The analysis was divided into two models, the first referring to the prediction of the number of deaths caused by COVID-19 and the second to the PSI prediction. In the first model, the original variables were used, while in the second model, a Principal Component Analysis was made, that were later used for the regression of the model. The method used for data processing was CRISP-DM. Data were obtained from an open access database. Finally, it should be noted that Gradient Boosting was the algorithm that obtained the best results according to the precision metrics that were used. It is worth highlighting the greater efficiency of the Ensemble and neural networks algorithms compared to the linear algorithms in the prediction of the data used.
Descrição
Mestrado Bolonha em Métodos Quantitativos para a Decisão Económica e Empresarial
Palavras-chave
COVID-19 óbitos PSI machine learning Portugal deaths
Contexto Educativo
Citação
Arriaga, Alexandre Poeiras (2022). “Impacto demográfico e financeiro da pandemia Covid 19 em Portugal : previsão do número de mortes e do PSI”. Dissertação de Mestrado. Universidade de Lisboa. Instituto Superior de Economia e Gestão
Editora
Instituto Superior de Economia e Gestão