A Hybrid Machine Learning Approach for Sleep and Awake States Detection Using EEG Data

Brain wave analysis significantly contributes in neurological diagnostics and research by clinically offering critical insights into neural activity and different patterns of sleep. This work aims in accurately detecting sleep stages using Electroencephalography (EEG) data using Exploratory Data Analysis (EDA) and advanced modelling techniques, with a focus on precisely identifying and categorizing sleep stages. The dataset for this study is extracted from Kaggle that includes complete EEG recordings of individuals during sleep is processed and visualized using EDA. During the modelling stage, predictive models for sleep stage classification are constructed using Machine Learning (ML) algorithms like Random Forest, SVM, KNN, and Logistic Regression. Using ensemble techniques, hyperparameter optimization, and cross-validation, these models go through rigorous training, validation, and testing. The most accurate model was a hybrid one that combined Random Forests, SVM, KNN, and Logistic Regression as a meta-learner. The polysomnographic recordings that make up the dataset provide rich brain wave data from various stages of sleep. The methodology was validated using benchmark datasets, which confirmed its high accuracy in identifying different stages of sleep. These results demonstrate the importance of rigorous data exploration and reliable modelling approaches for the advancement of neuroinformatics research.