Sleep Stage Classification Using Random Forest Method

The aim of this work is to apply Random Forest algorithm to classify REM and NREM sleep stages from a single-channel EEG. The training and performance evaluation of this classifier was performed on open-access data (Physiobank SLEEP-EDF database). A total of 5 features were extracted from 30 s epochs of non-overlapping windows. The proposed classifier has achieved an accuracy of 93.09% and Cohen's kappa of 0.90. The proposed classifier can be implemented on a portable microprocessing unit for in-home neuro-monitoring applications.