DATA-DRIVEN APPROACH FOR PHYSICAL ACTIVITY SEQUENCE ANALYSIS – CLUSTERING, PATTERN ANALYSIS AND PREDICTION OF SLEEP QUALITY

Abstract

Sleep is a vital human behavior that can affect the quality of life. Poor sleep quality can lead to a variety of health problems. Physical activity serves as a non-pharmaceutical intervention that improves sleep quality. Given the association between sleep and physical activity, there have been many efforts to investigate this relationship. Nowadays, wearables provide a non-intrusive solution for collecting sleep data. The ability to understand the correlation between sleep quality and daily physical activities provides an opportunity to improve humans’ quality of life._x000D_ _x000D_ This research study collected data from wearable devices. The raw signal from the wearable device is abstracted into a temporal state sequence of physical activities. These physical activity sequences are labeled with a quality of sleep. Once the data is cleaned and transformed, we apply an unsupervised machine learning for clustering similar activity sequences and supervised machine learning for predicting sleep quality given the physical activity sequences.

Date of Award	2019
Original language	American English
Awarding Institution	HBKU College of Science and Engineering