Random Forest Approach for Sleep Stage Classification

Abstract

• Sleep is known as a primary function and one of the most essential state of human brain which plays a vital role in both health and mental in individuals lives. Sleep stage classification has become an important factor in terms of sleep disease diagnosing and treatment. Brain Computer Interface (BCI) is a computer based analysis of brain signal which records as ElectroEncephaloGraphy (EEG) signals by using electrodes placed on the scalp. BCI has become a popular technology lately and sleep stage classification using this has become a research area over the last two decades. Statistical analysis was the base for all the difference classification techniques but with the evolution of machine learning, scientists and researchers has moved to machine learning sleep stage classification differencing feature selection and classification algorithms. Also different domain such as time domain, frequency domain and etc… were come across various studies. Random Forest is used ensemble techniques in machine learning. Random forest builds multiple decision trees than one to make its predication. This study focuses on building a Random Forest sleep stage classification with higher accuracy as 91%. A comparison has performed using three different classifiers as Support Vector Machine (SVM), Multi-Layer Perceptron and K-Nearest Neighbors. Further the effect of different channels in EEG signal has been represented in the study.