Capstone project — EEG and eye state

Overview

Integrative project on predicting eye state from EEG signals and a separate causal analysis. The work connects time-series preparation, machine learning, a Transformer Encoder and critical interpretation of results.

Technologies used

Key concepts covered

• Signal windowing and chronological splitting
• Comparison of tabular and sequential models
• ROC-AUC, balanced accuracy and macro F1
• Causal assumptions, estimated effects and uncertainty

Coursework and evaluated components

• Data preparation, training and evaluation notebook
• Comparison on 24 test windows; analysis of limitations
• Project report and results presentation