Online Program

Return to main conference page

All Times EDT

Friday, October 8
Fri, Oct 8, 1:15 PM - 2:30 PM
Speed Session

Discovering the Impact of Physical Activity Attributes on Sleep Using Machine Learning (309910)

*Alena J. Morrissette,  
Sreeranjani Ramprakash, Argonne National Laboratory 

Keywords: Computer Science, Machine Learning, Sleep, Physical Activity

Physical activity impacts a person’s sleep quality. Understanding the different activities that affect sleep will provide a deeper understanding on ways to improve the quality of sleep. This study focused on how physical activity attributes including lightly active minutes, moderately active minutes, very active minutes, and sedentary minutes impact the quality of sleep, using data from a Fitbit. The relationship between sleep quality and physical data was then modeled using the Random Forest method and used to predict which attribute has the most effect on the sleep score. We hypothesized that the minutes in the lightly active category would have the most impact on the quality of sleep, compared to the other attributes. The model results showed an accuracy of 0.97. This high number was due to a smaller sample size, and the estimated error rate was 20.92%. We determined the importance of the factors from the output of the Random Forest model using MeanDecreaseGini. The highest importance value was for the minutes spent being sedentary. This means that the amount of time someone spends being inactive has the greatest effect on their sleep quality, whether positive or negative.