12:00 - 14:00 | Thu 10 Nov | Maya Ballroom Foyer | ThPO.15
Sedentary behaviors such as prolonged occupational and leisure-time sitting are now ubiquitous in modern societies. Sedentary time is positively associated with increased risk of obesity, diabetes, cardiovascular disease, and all-cause mortality. Smartphones can sense the sedentary behaviors performed by their users, as well as the contexts (situations) in which sedentary behaviors occur. In this paper, we explore whether the contexts that can be sensed by users' smartphones can be used to predict their future sedentary behaviors reliably. We analyze data gathered in a term-long study of 49 college students in order to discover their sedentary behavior patterns and contexts strongly correlated with sedentary states. The ability to predict sedentary behaviors will facilitate more effective computer-driven interventions based on the theory of planned behavior. Using logistic regression, we are able to classify user context variables such as location, time, and app usage to predict if the user will be "very sedentary" in the next hour with a precision of 73.1% (recall of 87.7%).