Creative Commons

This dissertation was written as a part of the MSc in Data Science at the International Hellenic University. Sleep is a necessary part in every human’s life. Effective diagnosis and treatment of patients with sleep-related complaints is currently an urgent and heavily researched topic in the healthcare community. Sleep stage classification was introduced almost 50 years ago and manual approaches are sometimes used until today. Automatic sleep stage classification using machine learning can increase consistency and reliability, assisting experts at diagnosing sleep related health problems. In this dissertation, multiple classifiers are tested on 3 different datasets of healthy and patient subjects. The optimal algorithm achieves accuracy over 90% for the healthy subjects’ dataset. In the results, the difference between the EEG patterns of healthy and patient subjects is highlighted. It is finally concluded that, using mixed datasets from healthy people and patients with minor sleep disorders, decent classification accuracies can be achieved. In addition to that, algorithms can generalize better as they can be used for a larger number of people.