A1120
Title: Supervised fusion learning of physical activity features: Functional frameworks with L0 regularization
Authors: Margaret Banker - Northwestern Univeristy (United States) [presenting]
Abstract: Wearable devices are crucial in physical activity research because they provide continuous, real-time monitoring of various health metrics such as heart rate, physical activity, sleep patterns, and vital signs. These devices enable the collection of extensive, longitudinal data, offering insights into the daily lives and health trajectories of older adults. This information is invaluable for identifying early signs of health decline, assessing the effectiveness of interventions, and personalizing care plans. Wearable device data is considered in a functional framework with an L0 regularization approach, handling highly correlated micro-activity windows that serve as predictors in a scalar-on-function regression model. A longitudinal functional framework is developed with repeated wearable data to understand the influence of serially measured functional accelerometer data on longitudinal health outcomes. This method leverages quadratic inference function (QIF) via mixed integer optimization for longitudinal data analysis to detect critical physical activity windows and assess their population-average effects on health outcomes such as biological aging and body fat measurements.
