Method for estimation of weather and weekday effects on activity behavior acquired using wearable sensors

摘要

This paper presents the importance of association investigation between weather, weekdays and unsupervised physical activities of subjects. All data were collected using wristbands and smartphones. We used methods of descriptive statistics to derive activity streams such as periodicity, intensity, stable location, and velocity, and to find universal threshold values for segmentation of these streams automatically. We found that weather has an individual influence on activity patterns that can be used for personal profiling. Specifically, this influence varies significantly among subjects, depending on weekday factors.