Clustering and Classification Techniques for the Identification of Sleep Disorders Using the Sleep Health and Lifestyle of Worker

Abstract

Sleep disorders including Insomnia and Obstructive Sleep Apnea have a major impact on physical and psychological health and quality of life. Precise and timely detection of these diseases is essential for successful diagnosis and cure. Conventional expert-diagnosing the sleep disorder is based on manual interpretation of experience and it is very slow and subjective. As more lifestyle and health data can be retrieved, machine learning methods present an effective substitute in automatic detection of sleep disorder. In this paper, we present a comparative approach that evaluates unsupervised and supervised machine learning based sleep disorder discovery using the Sleep Health and Lifestyle Dataset. In the system presented, clustering methodologies - K-Means and DBSCAN - were initially used to find subtle pattern, and group people by sleep and lifestyle. Confusion matrices and classification reports evaluate clustering findings by associating cluster labels with sleep disorder classes. Furthermore, supervised classification algorithms are applied in the same dataset to directly predict sleep disorders. The performance of clustering-based methods and classification-based approaches has been considered by comparing them using the usual measures such as accuracy, precision, recall, F1-score and confusion matrices. The experimental analysis reveals the advantages and disadvantages of those approaches, showing that while clustering does find natural trends in the data, if labeled data are present, supervised classification methods can lead to a larger prediction.