Abstract:
Wearable activity trackers are electronic devices which are used to monitor fitness, sleep and other physical activities. The popularity of these devices is increasing as...Show MoreMetadata
Abstract:
Wearable activity trackers are electronic devices which are used to monitor fitness, sleep and other physical activities. The popularity of these devices is increasing as their capability to identify and track different type of activities has grown while at the same time they continue to become more affordable. This paper evaluates several machine learning techniques to improve our ability to accurately predict various physical activities like lying, sitting, walking and running using data tracked by such activity trackers. The dataset includes 3656 of Apple Watch and 2608 minutes of Fitbit data. The results show that an XGBoost model achieved the maximum AUC of 0.98 for the Apple Watch dataset, 0.99 on the Fitbit dataset, and 0.98 on the combined dataset.
Date of Conference: 27-29 August 2021
Date Added to IEEE Xplore: 04 October 2021
ISBN Information: