SavedAdd your journal

Journals / Fırat Üniversitesi Mühendislik Bilimleri Dergisi / 2026 / Cilt: 38 Sayı: 1

Predicting Stress Levels in University Students Using Machine Learning: An Optimization Approach Based on Synthetic Data

Üniversite Öğrencilerinde Stres Düzeyinin Makine Öğrenmesiyle Tahmini: Sentetik Veri Destekli Optimizasyon Yaklaşımı

Emre Canayaz(MARMARA ÜNİVERSİTESİ)0000-0002-3695-3642
Pages
285–298
Publication Date
2026-03-29
🏛️Bibliographic metadata and full text access for this article are provided via DergiPark.
DergiPark Record ↗
✨

AI Summary & Key Takeaways

AI Insight
⚡ TL;DR(Too Long; Didn't Read)

The results suggest that models supported by synthetic data and validated by explainable artificial intelligence are a reliable tool for early detection of students' stress levels.

📌 3 Key Takeaways

🎯
Objective & Hypothesis:University students are at high risk of stress due to academic pressure and social challenges.
📊
Key Findings & Data:This study aims to predict stress levels with machine learning using self-report-based lifestyle data from 2,000 students.
💡
Conclusion & Impact:The results suggest that models supported by synthetic data and validated by explainable artificial intelligence are a reliable tool for early detection of students' stress levels.
🔬Design:Modeling & Simulation
📍Scope:Comprehensive Sample
⚙️Variables:Stres düzeyi tahmini · makine öğrenmesi · sentetik veri

Abstract

University students are at high risk of stress due to academic pressure and social challenges. This study aims to predict stress levels with machine learning using self-report-based lifestyle data from 2,000 students. The class imbalance in the dataset was eliminated in the training phase with synthetic data generated by the Conditional Table Generating Generative Adversarial Network method, and the models were tested with stratified cross-validation. Among the five algorithms compared, the Random Forest model achieved the highest performance (Accuracy = 1.00; AUC = 1.00; Macro F1 = 1.00), correctly classifying all instances in both balanced and imbalanced data sets. The decision mechanism of the model was analyzed by Shapley analysis; “Daily Working Time” and “Sleeping Time” were found to be the most determining factors. The possibility that the high performance obtained was due to data leakage or random fit was supported by the fact that the independent test set was kept separate throughout the process and the findings of the ablation (feature extraction) and label permutation tests applied as robustness checks. The results suggest that models supported by synthetic data and validated by explainable artificial intelligence are a reliable tool for early detection of students' stress levels.

Özet

Üniversite öğrencileri, akademik baskı ve sosyal zorluklar nedeniyle yüksek stres riski taşımaktadır. Bu çalışma, 2.000 öğrenciden elde edilen öz-bildirim temelli yaşam tarzı verilerini kullanarak stres düzeylerini makine öğrenmesi ile tahmin etmeyi amaçlamaktadır. Veri setindeki sınıf dengesizliği, eğitim aşamasında Koşullu Tablo Üretici Üretken Karşıt Ağ yöntemiyle üretilen sentetik verilerle giderilmiş; modeller tabakalı çapraz doğrulama ile test edilmiştir. Karşılaştırılan beş algoritma arasında Rastgele Orman modeli hem dengeli hem dengesiz veri setlerinde tüm örnekleri doğru sınıflandırarak en yüksek performansa ulaşmıştır (Doğruluk = 1.00; AUC = 1.00; Makro F1 = 1.00). Modelin karar mekanizması Shapley analizi ile incelenmiş; "Günlük Çalışma Süresi" ve "Uyku Süresi" en belirleyici faktörler olarak saptanmıştır. Elde edilen yüksek başarımın veri sızıntısı ya da rastlantısal uyumdan kaynaklanma olasılığı, bağımsız sınama kümesinin süreç boyunca ayrı tutulması ve sağlamlık kontrolleri olarak uygulanan ablasyon (özellik çıkarma) ile etiket permütasyon testi bulguları ile desteklenmiştir. Sonuçlar, sentetik veriyle desteklenen ve açıklanabilir yapay zeka ile doğrulanan modellerin, öğrencilerin stres düzeylerinin erken tespitinde güvenilir bir araç olduğunu göstermektedir.

🔖

Cite This Article

Canayaz, E. (2026). Predicting Stress Levels in University Students Using Machine Learning: An Optimization Approach Based on Synthetic Data. Fırat Üniversitesi Mühendislik Bilimleri Dergisi, Cilt: 38 Sayı: 1, 285–298. https://doi.org/10.35234/fumbd.1805121

📄 Full Text PDF Viewer

📄

PDF görüntüleyici hazırlanıyor...

References (26)

  1. Lazarou E, Exarchos TP. Predicting stress levels using physiological data: Real-time stress prediction models utilizing wearable devices. AIMS neuroscience 2024; 11(2): 76.
  2. Kessler RC, Berglund P, Demler O, Jin R, Merikangas KR, Walters EE. Lifetime prevalence and age-of-onset distributions of DSM-IV disorders in the National Comorbidity Survey Replication. Archives of general psychiatry 2005; 62(6): 593-602.
  3. Xiong J, Lipsitz O, Nasri F, Lui LM, Gill H, Phan L, Chen-Li D, Iacobucci M, ve diğerleri. Impact of COVID-19 pandemic on mental health in the general population: A systematic review. Journal of affective disorders 2020; 277: 55-64.
  4. El Morr C, Jammal M, Bou-Hamad I, Hijazi S, Ayna D, Romani M, Hoteit R. Predictive machine learning models for assessing lebanese university students’ depression, anxiety, and stress during COVID-19. Journal of Primary Care & Community Health 2024; 15: 21501319241235588.
  5. Sieverding M, Schmidt LI, Obergfell J, Scheiter F. Stress und studienzufriedenheit bei bachelor-und diplom-psychologiestudierenden im vergleich. Psychologische Rundschau 2013; 64(2): 94-100.
  6. Ahuja R, Banga A. Mental stress detection in university students using machine learning algorithms. Procedia Computer Science 2019; 152: 349-53.
  7. Nayan MIH, Uddin MSG, Hossain MI, Alam MM, Zinnia MA, Haq I, Rahman MM, Ria R, ve diğerleri. Comparison of the performance of machine learning-based algorithms for predicting depression and anxiety among University Students in Bangladesh: A result of the first wave of the COVID-19 pandemic. Asian Journal of Social Health and Behavior 2022; 5(2): 75-84.
  8. Medikonda J. A clinical and technical methodological review on stress detection and sleep quality prediction in an academic environment. Computer methods and programs in biomedicine 2023; 235: 107521.
  9. Baba A, Bunji K. Prediction of mental health problem using annual student health survey: machine learning approach. JMIR Mental Health 2023; 10: e42420.
  10. Ratul IJ, Nishat MM, Faisal F, Sultana S, Ahmed A, Al Mamun MA. Analyzing perceived psychological and social stress of university students: A machine learning approach. Heliyon 2023; 9(6): e17004.
  11. Zhang J, Tee M, Lin C, Huili S. ZJ-EduFormer: Predicting Supply Chain Student Stress Using Transformer. In: 2024 5th International Conference on Information Science and Education (ICISE-IE); 2024; Rhodes, Greece. New York, NY, USA: IEEE. pp. 307-311.
  12. Đokić A, Stefanović H, Dudić D. Comparative analysis of classification model performance in predicting stress levels in students. In: Annual Conference on Challenges of Contemporary Higher Education (ACCHE); 3 February 2025; Belgrade, Serbia. pp. 61-66.
  13. Ghara A, Khan A, Mahata P. Lifestyle-Based Student Stress Detection with Real-Time ML Recommendations. International Journal of Innovative Research in Science, Engineering and Technology (IJIRSET) 2025; 14(6): 5.
  14. Bastos AF, Fernandes-Jr O, Liberal SP, Pires AJL, Lage LA, Grichtchouk O, Cardoso AR, de Oliveira L, ve diğerleri. Academic-related stressors predict depressive symptoms in graduate students: A machine learning study. Behavioural Brain Research 2025; 478: 115328.
  15. Tuan T, Loan D, Buddhahai B. Classification models combined with optimized features for mental stress prediction. International Journal of Data and Network Science 2025; 9: 737-50.
  16. Rahunathan L, A SN, G A, P R, S SL, N SV. Machine Learning Model for Student Stress Level Prediction. In: 2025 4th OPJU International Technology Conference (OTCON); 9-11 April 2025; Raigarh, India. pp. 1-7.
  17. Kumar S. Student Lifestyle Dataset: Daily Lifestyle and Academic Performance of Students. Kaggle; 2025.
  18. Xu L, Skoularidou M, Cuesta-Infante A, Veeramachaneni K. Modeling Tabular data using Conditional GAN. In: Advances in Neural Information Processing Systems 32; 8-14 December 2019; Vancouver, Canada. Red Hook, NY, USA: Curran Associates, Inc. pp. 7335-7345
  19. SDMetrics. SDMetrics 2025 [Available from: https://github.com/sdv-dev/SDMetrics].
  20. Esteban C, Hyland SL, Rätsch G. Real-valued (medical) time series generation with recurrent conditional GANs. arXiv preprint 2017; arXiv:1706.02633.
  21. Papadaki E, Vrahatis AG, Kotsiantis S. Exploring innovative approaches to synthetic tabular data generation. Electronics 2024; 13(10): 1965.
  22. Williamson BD, Gilbert PB, Simon NR, Carone M. A general framework for inference on algorithm-agnostic variable importance. Journal of the American Statistical Association 2023; 118(543): 1645-58.
  23. Ojala M, Garriga GC. Permutation tests for studying classifier performance. J Mach Learn Res 2010; 11: 1833-1863.
  24. Tran D-S, Nguyen D-T, Nguyen T-H, Tran C-T-P, Duong-Quy S, Nguyen T-H. Stress and sleep quality in medical students: a cross-sectional study from Vietnam. Frontiers in psychiatry 2023; 14: 1297605.
  25. Ban DJ, Lee TJ. Sleep duration, subjective sleep disturbances and associated factors among university students in Korea. Journal of Korean Medical Science 2001; 16(4): 475.
  26. Bozorgmehr A, Weltermann B. Prediction of chronic stress and protective factors in adults: development of an interpretable prediction model based on XGBoost and SHAP using national cross-sectional DEGS1 data. JMIR AI 2023; 2: e41868.