Sleep Monitoring with Wearable Sensor Data in an eCoach Recommendation System: A Conceptual Study with Machine Learning Approach

The collective effects of sleep loss and sleep disorders are correlated with many adverse health consequences, including increased risk of high blood pressure, obesity, diabetes, depressive state, and cardiovascular symptoms. Research in eHealth can provide methods to enrich personal health care with information and communication technologies (ICTs). An eCoach system may allow people to manage a healthy lifestyle with extended health state monitoring (e.g., sleep) and tailored recommendation generation. Using supervised machine learning (ML) techniques, this study investigated the possibility of classifying sleep stages at night for adults on hourly and daily basis. The daily total sleep minutes and hourly total sleep minutes for defined sleeping period served as input for the classification models. We first used publicly available Fitbit dataset to build the initial classification models. Second, using the transfer learning approach, we re-used the top five best-performing models on a real dataset as collected from the MOX2-5 wearable medical-grade activity device. We found that support vector classifier (SVC) with “linear” kernel outdated other classifiers with a mean accuracy score of 99.92% for hourly sleep classification and a K-nearest neighbor (KNN) outpaced other classifiers with a mean accuracy score of 99.47% for daily sleep classification, for the public Fitbit datasets. Moreover, to determine the practical efficacy of the classifier models, we conceptualized to use the classifier models in an eCoach prototype system to attain tailored sleep goals (e.g., a weekly goal of 49–63 h of sleeping).