Nabarun Halder, Armun Alam, Jahanggir Hossain Setu, Asif Mahmood, Ashraful Islam, M Ashraful Amin
2025 17th International Conference on Knowledge and Smart Technology (KST)
In: 2025 17th International Conference on Knowledge and Smart Technology (KST)
IEEE, pp. 104-109
Stress is a psychological strain with diverse causes that can significantly impact health, potentially leading to conditions such as depression, heart disease, and impaired coping abilities if prolonged. While excessive stress is often harmful, moderate levels, known as eustress, can enhance motivation and performance. This study uses a publicly available dataset containing responses from 843 participants, aged 14 to 100, who rated their stress levels as distress, eustress, or no stress based on lifestyle and personal factors such as academic workload, social connections, and self-confidence. To classify these stress types, six machine learning models were developed: Logistic Regression (LR), Random Forest (RF), Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Decision Tree (DT), and a Multilayer Perceptron (MLP). During pre-processing, we identified a significant class imbalance, which could skew the results. To address this, the Synthetic Minority Oversampling Technique (SMOTE) was applied, ensuring a more balanced dataset and improving model reliability for underrepresented classes. Among the models, the MLP, specifically the Ablation${ }_{2}$architecture, achieved the highest performance, with an accuracy of 97.44 % and an F1-score of 97.36 % on the resampled dataset. While most models showed improved results after resampling, KNN's performance declined from 94.44 % to$\mathbf{7 7. 7 8 \%}$. This study highlights the importance of data balancing and model architecture in accurately classifying stress levels.
Assistant Professor
Department of Computer Science and Engineering
Independent University, Bangladesh (IUB)Human-Computer Interaction, AI for Social Good, AI for Public Health, AI for Impact
Machine Learning, Cognitive & Vision Science, Cybernetics, Surveillance & Security, ICT in Education, Health, & Agriculture, Human-Computer Interaction, Internet of Things, Robotics