Project Big Data - Group 22 Fitbit Analysis
Yildiz, Damla Akaçık, Onder Giaj Levra, Federico
3 July 2022
Introduction
In this project, we are using Fitbit Dataset which is a generated by 30 participant’s personal tracker data, including minute-level output for physical activity, heart rate, and sleep monitoring [3][4]. We will try to answer our research questions which covers a wide range, from sleeping habits of the participants to hours participants are most active, and thus helps us analyze the dataset in a systematic way. In the end, we will try to draw conclusions from these analysis and use them to fit regression models to our dataset.
Conclusion
Different conclusions for each question investigated were drawn. Those conclusions will be summed up in this part.
In the first part, activity of users was investigated by three main questions. For the first question, it was shown that there is a linear relationship between the number of steps taken and calories burned. Also, BMR value for the proper data points was investigated. The data was coming from 12 different users and according to the mean, standard deviation, and range of the BMR values, it was deduced that those 12 people might be mostly men. For the second question, it was concluded that there is a correlation between the type of activity and the average speed for that category. More active the type of activity, the greater the average speed. Investigating the third question, users were clustered into two groups, more sedentary or not. Also, it was concluded that for the less sedentary group, users tend to be more sedentary on weekdays. Lastly, it was observed that the average total sleeping time for the less sedentary group is greater than the average total sleeping time for the more sedentary group.
In the second part, we were able to inspect the sleep and activity habits of fitbit users. We discovered that many users only 5 out of 21 users are getting an amount of sleep which is recom- mended by the sleep foundation. Moreover, good sleepers are slightly more active than those who don’t sleep well. But the greatest difference lays in the kind of activity performed. In fact, good sleepers appear to be more constant in their activity, as shown by the packed interquartile ranges, and also less prone to burn a large amount of calories. This analyses ended with the suggestion that those which would like to improve their sleep habits might try to adopt an activity profile as the good sleepers, by being slightly more active during the day but without large energy efforts.
In the last part, we first wanted to see whether we can use the heart rate data to identify health problems among the users but quickly realized it would not be reliable due to lack of data for the age, gender, stress levels of the participants. Then we wanted to find out if there are any patterns on the days or hours people prefers to exercise and saw that there is no correlation between the day in a week and amount of activity done on that day. On the other hand, it was clear that participants preferred to exercise on certain times of the day particularly they were most active between 16:00- 18:00. Lastly, we investigated the relationship between different features to select the ones we will use in steps prediction. We tried 3 different models but in the end using ridge regularization gave us the best result.
Bibliography
- American heart association. https://www.heart.org/en/healthy-living/fitness/ fitness-basics/target-heart-rates, Mar 2021.
- How much sleep do we really need? https://www.sleepfoundation.org/how-sleep-works/ how-much-sleep-do-we-really-need#:~:text=National%20Sleep%20Foundation% 20guidelines1,to%208%20hours%20per%20night., Apr 2022.
- Robert Furberg, Julia Brinton, Michael Keating, and Alexa Ortiz. Crowd-sourced Fitbit datasets 03.12.2016-05.12.2016. https://doi.org/10.5281/zenodo.53894, May 2016.
- MOBIUS.¨ Fitbit fitness tracker data. https://www.kaggle.com/datasets/arashnic/fitbit, 2020.
- Bhupesh Panchal. What is basal metabolic rate (BMR)? https://www.hollandandbarrett. com/the-health-hub/weight-management/fitness/exercise/what-is-bmr/, 2021. 15