Automatic Classification of Insomnia Using Machine Learning Algorithms From EEG Signals

Insomnia is a common sleep disorder characterized by difficulty falling asleep or staying asleep, even when the sleep environment is conducive to rest. It can negatively affect both physical and mental health. It is often triggered by stress, anxiety, or underlying health conditions and impacts a large portion of the global population. This study presents an automated approach for detecting insomnia through EEG signals. The EEG data were subjected to multiple preprocessing steps, such as downsampling, bandpass filtering, artifact removal, and segmentation. The features were extracted from frequency bands of signals using the Fast Fourier Transform (FFT), and Recursive Feature Elimination (RFE) was applied to decrease the features from 70 to 30, enhancing classification accuracy. A total of $\mathbf{1 5, 4 8 0}$ instances were obtained for 22 subjects. After that, various machine learning models, including Random Forest, Decision Tree, k-Nearest Neighbor (kNN), and XGBoost, were applied for classification. Among them, the XGBoost model achieved the highest accuracy of 99.41 %. The results demonstrate the potential of EEG-based machine learning techniques for accurate insomnia detection, offering a promising method for clinical applications.