Visualizing Optimal Classifiers in EEG-Based Sleepy Driver Prediction

Driver drowsiness has long been associated with road accidents, often resulting in injuries and fatalities. Recognizing the significance of mitigating this issue, this study utilizes EEG data to predict drowsiness. This paper presents a comprehensive approach to EEG-based driver drowsiness prediction, addressing a critical road safety concern. Driver drowsiness has long been associated with road accidents, often resulting in injuries and fatalities. Recognizing the significance of mitigating this issue, this study utilizes EEG data to predict drowsiness. Electroencephalogram (EEG) is a non-invasive technique used to monitor alertness levels in individuals by recording brain activity. With a dataset related to EEG signals, the study tackles class imbalance using the Synthetic Minority Over-sampling Technique (SMOTE). Beyond data preprocessing, it explores data visualization and employs diverse machine learning models, aiming to uncover classification method strengths and weaknesses. Ensemble methods like Stacking and Voting achieve an impressive 82.69% test accuracy, with Extra Trees excelling in precision, Logistic Regression in recall, and CatBoost maintaining balanced performance. Leveraging PyCaret, this research advances EEG-based drowsiness detection, forming a robust foundation for improved drowsiness detection systems, ultimately enhancing road safety by facilitating more effective drowsiness identification and guiding future research.