Presentation
Investigating College Student Stress Using Smartwatch Accelerometer Data
Event Type
Lecture
Student Forum
TimeFriday, October 14th9:30am - 9:45am EDT
LocationA708
DescriptionStress is a major issue, especially among college students. Measuring human psychological dynamics, such as stress, is difficult because of the subjective nature of self-reporting and variability between and within individuals. Smartwatches facilitate objective measurement of behavior associated with psychological state and stress. This study investigated the possibility of detecting behavior that correlates with stress using only built-in smartwatch accelerometer sensor data. We collected acceleration data within six weeks from 45 undergraduate and graduate students from a large university in Texas. Subjects also self-reported perceived stress on their phones. Three model schemes were used to train the learning algorithms for stress detection. We achieved a maximum accuracy of 88.91% for the user-specific model, 76.83% for the general model, and 84.41% for the similar-users model. Results indicate that the proposed methodology can be a potential system for stress recognition relying solely on data from a wrist-worn accelerometer.
