Publication: Yapay zeka tabanlı uyku analizi ve evre sınıflandırması
Date
Authors
Authors
Başaran, Ömer Faruk
Advisor
Semerci, Neyir Özcan
Language
Type
Publisher:
Bursa Uludağ Üniversitesi
Journal Title
Journal ISSN
Volume Title
Abstract
Bu tez çalışmasında, literatürde yaygın olarak kullanılan Physionet Sleep-EDF(Expanded) EEG veri seti ile 6 uyku ve uyanıklık evresinin sınıflandırılması yapılmıştır. Veri seti incelenmiş sınıflandırmak için uygun denekler ve kayıtlar seçilmiştir. Veri setinde oluşan dengesizlikler çözülmeye çalışılmış ve sonrasında diğer aşamalara geçilmiştir. Önişleme aşamasında, EEG sinyalleri delta, teta, alfa, beta ve gama frekans bantlarına göre Butterworth IIR filtreleri ile işlenmiş ve epoklara ayrılmıştır. Öznitelik çıkarımı aşamasında, PSD ve Esis başta olmak üzere birçok öznitelik çıkarılmıştır ve çıkarılan öznitelikler başarıma etkilerine göre karşılaştırılmıştır. Sınıflandırma gerçekleştirilirken ise Random Forest, Gradient Boosting gibi çeşitli algoritmalar test edilmiş ve başarılarına göre karşılaştırılmışlardır.
In this thesis study, the classification of six sleep and wakefulness stages was conducted using the Physionet Sleep-EDF (Expanded) EEG dataset, which is widely used in the literature. The dataset was examined, and appropriate subjects and recordings were selected for classification. Efforts were made to resolve any imbalances present in the dataset before proceeding to other stages. During the preprocessing stage, EEG signals were processed and segmented into epochs using Butterworth IIR filters based on delta, theta, alpha, beta, and gamma frequency bands. In the feature extraction stage, various features, including PSD and Esis, were extracted and compared based on their impact on performance. For the classification, algorithms such as Random Forest and Gradient Boosting were tested and compared based on their performance.
In this thesis study, the classification of six sleep and wakefulness stages was conducted using the Physionet Sleep-EDF (Expanded) EEG dataset, which is widely used in the literature. The dataset was examined, and appropriate subjects and recordings were selected for classification. Efforts were made to resolve any imbalances present in the dataset before proceeding to other stages. During the preprocessing stage, EEG signals were processed and segmented into epochs using Butterworth IIR filters based on delta, theta, alpha, beta, and gamma frequency bands. In the feature extraction stage, various features, including PSD and Esis, were extracted and compared based on their impact on performance. For the classification, algorithms such as Random Forest and Gradient Boosting were tested and compared based on their performance.
Description
Source:
Keywords:
Keywords
Makine öğrenmesi, Sınıflandırma, Gradient boosting, EEG, Uyku, Uyku evreleri, Yapay zeka, Artificial intelligence, Machine learning, Classification, Sleep, Sleep stages