Automated Sleepiness Detection via EEG Brainwave Analysis: A Nonlinear Ensemble Approach with Optimized Hyper-Tuning Strategies

Abstract

Introduction: Automated sleepiness detection plays a vital role in ensuring driving-related safety, monitoring physical well-being, worker safety, and various other applications. The ability to detect and respond to signs of sleepiness is essential in preventing accidents and improving overall safety and productivity.

Objectives: The primary objective of this work was to explore avenues for enhancing the performance of automated sleepiness detection systems using advanced machine learning techniques. Specifically, the study aimed to determine the improvement potential in accuracy through the application of fine-tuning techniques across several machine learning algorithms.

Methods: A range of machine learning algorithms, including LightGBM, XGBoost, CatBoost, Extra Trees Classifier (ETC), and Random Forest (RF), were employed to evaluate their effectiveness in sleepiness detection. Fine-tuning techniques were applied to these boosting algorithms to assess their impact on performance improvements.

Results: The results revealed significant performance enhancements, particularly for boosting algorithms. CatBoost emerged as the top performer, achieving an accuracy score of 83%, demonstrating its capability in the domain of sleepiness detection. The improvements achieved through fine-tuning were substantial across all algorithms evaluated.

Conclusions: This study highlights the potential of advanced machine learning techniques to make meaningful contributions to automated sleepiness detection systems. The results underscore the importance of fine-tuning in boosting algorithm performance and suggest a growing role for such technologies in improving safety, health, and productivity in various domains.