<b>Machine Learning</b>

This study explored the application of eye tracker technology in visual fatigue assessment, aiming to classify and predict fatigue status through nonlinear parameters. Visual fatigue is a common problem during long-term visual tasks, which has a great impact on personal health and work efficiency. In order to effectively evaluate visual fatigue, this study used eye movement data and calculated nonlinear parameters such as entropy, combined with a machine learning model for fatigue classification. In terms of methodology, we used three machine learning models: SVM, KNN, and Random Forest for comparison, and trained and verified the models based on eye movement parameters and SSQ questionnaire results. The results show that the Random Forest model performs best in classifying fatigue status with an accuracy of 0.76 and an AUC score of 0.81. The best parameter combination includes distance to center of mass(Dist_CM), fixaton count with the 10% AOI area(FC_10), and right eye Sample Entropy(RE_Entropy), showing that these parameters can effectively reflect the changes in fatigue status. The study confirmed the key role of nonlinear parameters such as entropy in the classification of visual fatigue and highlighted the potential of eye tracker technology in fatigue monitoring and management.