Exploratory research | Machine learning | Physiological signals
An exploratory project examining physiological features and machine-learning approaches for identifying states associated with elevated error risk in high-stakes settings.
The repository contains feature tables, evaluation outputs and figures. It is a research artefact, not a clinically validated system or operational safety product.
| Directory | Contents |
|---|---|
features/ |
Prepared feature tables, including features_all_subjects.csv and features_drozy.csv |
results/ |
Model evaluation summaries, including leave-one-subject-out (LOSO) result tables |
figures/ |
ROC curves, per-subject AUC, feature importance, confusion matrix and conceptual intervention framework |
The available materials document intermediate analyses and visual outputs. A complete reproducible training pipeline, dataset provenance notes and formal validation protocol are not included in this repository.
The figures and results should be treated as exploratory modelling outputs. They should not be interpreted as evidence of reliable real-world error prediction or suitability for safety-critical deployment.
Shamseldin Abdelrahim