EEG-Based Mental State Prediction Using Hybrid Machine Learning Techniques

Abstract

This paper presents a Brain-Computer Interface (BCI) system that reads EEG signals and classifies a person's internal state as either Concentrated or Relaxed. A complete pipeline is constructed covering raw EEG data loading to predictions using machine learning. The system uses the MNE-Python library to read EDF files, applies a bandpass filter (8–30 Hz), and extracts frequency band power features from five standard EEG bands: Delta, Theta, Alpha, Beta, and Gamma. These features are used to train a Random Forest classifier. Four different models — Random Forest, SVM, LSTM, and 1D CNN — are compared to evaluate traditional machine learning against deep learning on this time-series classification problem. Results show that Random Forest achieves 93–96% accuracy on the standard task, and deep learning models reach similar levels. The work provides an open-source, reproducible framework suitable for EEG research in healthcare, education, and neuroscience.