Detecting Students Stress and Fatigue in Online Learning using Machine Learning
B. Moganapriya, F. Mary Harin Fernandez · 2025
The growing trend towards online education has created concerns regarding students' stress and fatigue, which have negative effects on academic achievement and mental well-being. Conventional methods of identifying stress, including surveys and self-reporting, are subjective and not in real-time. This paper suggests a new DL-based method employing a hybrid CNN-LSTM model to identify levels of stress and fatigue among students. The proposed model incorporates multimodal information, such as facial expressions, voice tone, and keystroke dynamics, to improve detection accuracy. The designed framework analyzes webcam-recorded emotions, speech, and typing patterns to classify stress levels accurately. This trained and tested our model on data gathered from university students enrolling in online classes, with 92.3% accuracy, outperforming conventional machine learning techniques such as SVM and Random Forest. This study offers a basis for real-time monitoring systems in online learning, facilitating timely interventions to enhance student well-being and learning achievement. Future research can integrate physiological sensor data for increased accuracy.