User Exercise Pattern Prediction through Mobile Sensing

Abstract

Even though the health benefits of regular exercising are well known, an average person has difficulty maintaining physical activity on a regular basis. One of the main reasons for this is lack of motivation. With their increasing ubiquity, wireless devices and smartphones and their sensing capabilities now can be involved in solving this issue. Many mobile applications have been developed with which people are able to keep track of their exercises, become more aware of their physical condition, and be more motivated. The collected data is also a good source for researchers in understanding the exercise patterns and the main factors influencing people to exercise. Understanding those factors will allow better applications to be built, which helps motivate people. In this work, we quantitatively analyze a dataset collected from over 10,000 users. To better understand the user exercise patterns, we identify a set of factors influencing their exercise patterns. Based on these insights, we develop a prediction model to predict users' future exercise activities.

Recommender System for Responsive Engagement of Senior Adults in Daily Activities

“…Much work has focused on early prediction of future health behavior that could be used in preventive healthcare. These works include: predicting future cognitive impairment in elderly people from variables, which are commonly collected in community health care institutions (Na 2019), predicting mortality in elderly people from medical history, diet, exercises and lifestyle activity (Leitzmann et al 2007), predicting mortality in older women from mean daily step counts (Lee et al 2019), predicting changes in exercise behavior from historical physical activity data (Kotsev et al 2014), predicting daily blood pressure levels from historical blood pressure and health behavior (Chiang and Dey 2018), predicting in-hospital mortality, readmission, prolonged length of stay and final discharge diagnoses from electronic health records data (Rajkomar et al 2018), predict future actions from past activities (Kurashima et al 2018), etc. The early prediction information can be useful for a health recommender system to decide when it needs to act, however, it is not sufficient to decide how to act.…”