| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 1.64 MB | Adobe PDF |
Abstract(s)
O peso corporal é muito mais que um número numa balança. Este valor pode ser
indicador de várias doenças, pois tanto o excesso como a falta de peso têm implicações na saúde dos indivíduos. O excesso de peso está associado a doenças cardíacas,
obesidade, diabetes, hipertensão arterial, distúrbios respiratórios, entre outras. Enquanto a falta de peso num nível extremo está associado a problemas de deficiência
nutricional, enfraquecimento do sistema imunológico, osteoporose e desequilíbrios
hormonais. Devido a estes problemas, surge a necessidade de acompanhar e analisar
as alterações corporais, para a adoção de uma dieta e um estilo de vida equilibrado
com as necessidades do indivíduo.
O processo de controlo do peso é um processo complicado e está dependente de
vários fatores. Assim sendo, e considerando que a versatilidade da área de Machine
Learning (ML) permite desenvolver projetos que melhorem a qualidade de vida do
ser humano, neste trabalho pretende-se desenvolver um modelo de ML para prever
o peso futuro tendo em conta o registo alimentar, exercício físico e Taxa Metabólica
Basal (TMB) de indivíduo, com o objetivo de mostrar o impacto que três dias podem
ter no peso futuro.
Os resultados da performance do modelo obtidos através do cálculo das métricas
de desempenho, foram positivos. Através do cálculo do Coeficiente de Determinação
foi obtido o valor 0.75, o que para esta métrica é considerado um valor bom, visto
que está mais próximo de 1 do que de 0. Os valores do cálculo do Mean Squared
Error (MSE) e do Mean Absolute Erro (MAE) demonstra que o modelo conseguiu
aprender os padrões nos dados e que não existiu overfitting significativo. Estes
resultados demonstram ser viável o desenvolvimento deste tipo de soluções.
ABSTRACT: Body weight is much more than just a number on a scale. This value can be indicative of various diseases, as both excess and insufficient weight have implications for individual’s health. Excess weight is associated with heart disease, obesity, diabetes, high blood pressure, respiratory disorders, among others. Meanwhile, extreme underweight is associated with problems such as nutritional deficiency, weakened immune system, osteoporosis, and hormonal imbalances. Due to these issues, there is a need to monitor and analyze body changes in order to adopt a diet and lifestyle balanced with individual needs. The process of weight control is a complicated one and depends on various factors. Therefore, considering the versatility of the field of ML, which allows for the development of projects that improve human quality of life, this study aims to develop a ML model to predict future weight based on dietary records, physical exercise, and TMB, with the goal of demonstrating the impact that three days can have on future weight. The results of the model’s performance obtained through the calculation of performance metrics were positive. The coefficient of determination yielded a value of 0.75, which for this metric is considered good, as it is closer to 1 than to 0. The values of MSE and MAE demonstrate that the model was able to learn patterns in the data and that there was no significant overfitting. These results indicate the viability of developing such solutions.
ABSTRACT: Body weight is much more than just a number on a scale. This value can be indicative of various diseases, as both excess and insufficient weight have implications for individual’s health. Excess weight is associated with heart disease, obesity, diabetes, high blood pressure, respiratory disorders, among others. Meanwhile, extreme underweight is associated with problems such as nutritional deficiency, weakened immune system, osteoporosis, and hormonal imbalances. Due to these issues, there is a need to monitor and analyze body changes in order to adopt a diet and lifestyle balanced with individual needs. The process of weight control is a complicated one and depends on various factors. Therefore, considering the versatility of the field of ML, which allows for the development of projects that improve human quality of life, this study aims to develop a ML model to predict future weight based on dietary records, physical exercise, and TMB, with the goal of demonstrating the impact that three days can have on future weight. The results of the model’s performance obtained through the calculation of performance metrics were positive. The coefficient of determination yielded a value of 0.75, which for this metric is considered good, as it is closer to 1 than to 0. The values of MSE and MAE demonstrate that the model was able to learn patterns in the data and that there was no significant overfitting. These results indicate the viability of developing such solutions.
Description
Keywords
Machine Learning Redes neuronais Peso Corporal Taxa metabólica basal Equação de Harris e Bennedit