EEG-BASED ANXIETY AND STRESS DETECTION USING PCA-RFE HYBRID FEATURE EXTRACTION AND XGFOREST CLASSIFICATION WITH HYPERPARAMETER OPTIMIZATION

Abstract

Modern research about mental health analyses Electroencephalography (EEG) data for detecting and categorizing anxiety and stress manifestations. The research presents an improved detection method which unites innovative features extraction methods with optimization approaches as well as advanced classification techniques to enhance accuracy. Our proposed method known as PCA-RFE Hybrid Feature Extraction implements Principal Component Analysis to reduce dimensions alongside Recursive Feature Elimination for selecting important EEG features thus achieving better model interpretability coupled with improved computational speed. The XGForest Classifier stands as our classification model because it unites XGBoost and Random Forest algorithms to maximize predictive accuracy at 93%. The best parameters are chosen from GridSearchCV to achieve maximum classification accuracy during hyperparameter optimization. The Hybrid Classifier obtains 1.00 accuracy along with precision 0.99, recall 0.99 and F1-score 0.97. The investigative findings indicate that the implemented method exhibits superior diagnostic performance regarding traditional machine learning algorithms because XGBoost reached 0.93 accuracy while Neural Network achieved a dismal 0.37 accuracy. Early detection of mental health problems becomes possible through the development of this study which works toward creating improved EEG-based diagnostic tools for monitoring purposes.