Baseline Model for EEG Classification using MiniROCKET + Ridge Classifier

This project implements a baseline model for EEG signal classification using:

• MiniROCKET (for time series feature extraction)
• RidgeClassifierCV (for classification)

Dataset:

• BCI Competition IV Dataset 2a (BNCI2014_001) via MOABB.
• Using only Subject 1, Session '0train', Run '0' for this baseline.

Methodology:

1. Load raw EEG signals.
2. Extract event markers (motor imagery tasks: feet, left hand, right hand, tongue).
3. Epoch signals into trials (0–2 seconds windows).
4. Transform using MiniROCKET.
5. Train RidgeClassifierCV.
6. Evaluate accuracy & classification metrics.

Results (Baseline):

Metric	Value
Accuracy	~50%
Macro F1	~0.47
Support	10 samples (small test set)

Note: Accuracy is low because only a small subset (run '0' of subject 1) was used. This is intentional to establish a reproducible baseline.

Future Improvements:

• Combine multiple runs per subject (e.g., runs '1', '2', '3', etc.).
• Include more subjects (1–9) for better generalization.
• Experiment with S-Rocket, HDC-Rocket as enhancements.
• Evaluate on larger test splits & cross-validation.

Usage:

pip install -r requirements.txt
python src/data_preprocessing.py