Prediction of oxygen uptake (VO2) using neural networks
Ashfaq, Atiqa (2022)
2022
Master's Programme in Computing Sciences
Informaatioteknologian ja viestinnän tiedekunta - Faculty of Information Technology and Communication Sciences
This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Hyväksymispäivämäärä
2022-04-12
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:tuni-202204063079
https://urn.fi/URN:NBN:fi:tuni-202204063079
Tiivistelmä
This thesis focuses on using neural network models for the prediction of oxygen uptake (VO2). The predictions are made using regression techniques. The dataset contains independent predictor variables such as heart rate (HR), energy expenditure (EE), height, body mass, gender and age. VO2 is the output dependent variable. The goal is to evaluate and compare the performance of neural networks to other machine learning techniques such as support vector machines and multiple linear regression.
Few neural network models have been tested previously in the literature for maximal oxygen uptake (VO2max) prediction. During the last decade, most approaches have focused on support vector machines and linear regression equations. In this thesis, data collected at the University of Jyväskylä is used to create a dataset for the prediction of VO2. A detailed statistical analysis has been performed to see the relationship between speed, VO2 and energy expenditure. Using 8 different combinations of predictor variables, neural network’s performance and the effect of predictor variables on the performance is measured. Data pre-processing is performed. R2 value and root mean square error value is used for measuring the performance of the machine learning models. Same data set is used for all models to ensure accurate results.
The results of this thesis show that speed, VO2 and energy expenditure have a direct relationship. Males show higher energy produced as compared to females. The neural network model outperformed support vector machine and multiple linear regression by resulting in accurate predictions, high R2 value and low root mean square value. The highest accuracy is achieved with the model containing all predictor variables. The inclusion of HR as a predictor variable is important due to its effect on the performance of the model.
Further advancements in neural networks can allow more accurate VO2 predictions, the model can also be used in a wearable device for real-time VO2 prediction. The same approach can be extended to predict VO2max values.
Few neural network models have been tested previously in the literature for maximal oxygen uptake (VO2max) prediction. During the last decade, most approaches have focused on support vector machines and linear regression equations. In this thesis, data collected at the University of Jyväskylä is used to create a dataset for the prediction of VO2. A detailed statistical analysis has been performed to see the relationship between speed, VO2 and energy expenditure. Using 8 different combinations of predictor variables, neural network’s performance and the effect of predictor variables on the performance is measured. Data pre-processing is performed. R2 value and root mean square error value is used for measuring the performance of the machine learning models. Same data set is used for all models to ensure accurate results.
The results of this thesis show that speed, VO2 and energy expenditure have a direct relationship. Males show higher energy produced as compared to females. The neural network model outperformed support vector machine and multiple linear regression by resulting in accurate predictions, high R2 value and low root mean square value. The highest accuracy is achieved with the model containing all predictor variables. The inclusion of HR as a predictor variable is important due to its effect on the performance of the model.
Further advancements in neural networks can allow more accurate VO2 predictions, the model can also be used in a wearable device for real-time VO2 prediction. The same approach can be extended to predict VO2max values.