Empowering Wearable Sensor Generated Data to Predict Changes in Individual’s Sleep Quality Wahyu Hidayat 1 , Toufan D. Tambunan 2 , Reza Budiawan 3 1,2,3 School of Applied Science Telkom University Bandung, Indonesia wahyuhidayat@telkomuniversity.ac.id 1 , tambunan@telkomuniversity.ac.id 2 , rezabudiawan@telkomuniversity.ac.id 3 Abstract—Wearable sensors found in popular wrist wearable device are both generating sales profit and constantly generating vast amount of data. Some of these wearable sensors are able to record physical activity and sleep trends, both are being mainly used to give insight to its users about their current and past health and well-being. We proposed a method of data pre- processing and machine learning using simple k-nearest neighbor classifier to furthermore empower the usage of such data to predict changes in one’s sleep quality based on his or her current physical activity level. Our method were challenged to predict changes in five medically-approved sleep quality indicators, using data generated by commercially available consumer-grade wrist wearable device. The experiment result shows that the successful prediction of changes in sleep quality using wearable sensor generated data can be achieved by successfully selecting and sometimes combining the right input parameter(s). Each sleep quality indicators calls for different input parameter or combined parameters. By selecting and combining the right parameter(s), our method had successfully predict changes in both sleep duration and sleep efficiency with accuracy of 68% and 64%, respectively. Keywords—wearable;machine learning;k nearest neighbor I. INTRODUCTION Wearable sensors are commonly found in wrist wearable device product sold worldwide. Smartband, smartwatch and activity tracker are some of the most popular wrist wearable device. These wrist wearable device product are both popular and highly profitable. About 232 million units smartwatch has been sold worldwide at 2016, generating 28.7 billion USD sales profit, among 11.5 billion were from smartwatch sales profit alone [1]. Various sensors are embedded on these devices. A high-end device may contains few sensors, i.e. accelerometer, altimeter, optical heart rate sensor, bioimpedal sensor and body temperature sensor. Most common device will contain at least accelerometer and optical heart rate sensor. Each sensor constantly generate its own data which can be combined with user’s personal data to give information to user. For example, accelerometer generate user’s movement data that later can be combined with user’s gender, body weight, body height and age to estimate steps taken, distance covered and calorie burnt. If optical heart rate sensor also present then by combining data generated by accelerometer and optical heart rate sensors, a more accurate sleep record will also be available to user [2]. At present, the main usage of these data are limited to monitor or to give insight to users about his or her current and past health and well-being. Hypothetically, by collecting the same data over time and employing some machine learning algorithm, there is a far more potential usage of these data, that is to predict some aspect of his or her health. Hence, the benefit of these data can be elevated, for users not only presented with information but also prediction. Thus, by combining prediction with current status information, user can have a personal recommendation to take action based on desired output. This paper presents a method of data pre-processing and machine learning that utilize wearable sensor generated data to predict changes in individual’s sleep quality, that indicated by the change in one of five sleep quality indicator. Among those sleep quality indicators are sleep duration, sleep efficiency and sleep stage duration in all three stages of sleep; REM, light and deep sleep. The remaining of this paper is organized as follow: Section 2 presents a comparative study of current related works on wearable sensor generated data. In Section 3, the proposed method are explained in detail. Section 4 presents test metrics, scenario, and result. Finally in Section 5 we present conclusion and future works to improve this method. II. RELATED WORKS Numerous previous research that utilize wearable technology has been done. Wearable device were used to monitor heart rate and blood pressure for patients in remote areas that lack access to decent health infrastructure [3] and to monitor elderly patients that has chronic disease [4]. Some research observe wearable device’s security aspect [5] or privacy aspects [6]. Meanwhile, research on sleep study seek correlation between physical activities and sleep quality. For example, [7] tried to identify correlation between physical activity, gadget playing habit and sleep quality while [8] tried to identify correlation between diurnal blood pressure, physical activity and sleep quality. Some of well accepted and widely used