An Explainable Machine Learning Framework Analyzing Mental Health Influencing Metrics from Screen Time Behavior

This study explores the impact of screen-time-related behavioural factors on children's mental health using machine learning and explainable artificial intelligence techniques. Initially, Random Forest and CatBoost models are compared using multiple evaluation metrics, including accuracy, precision, recall, and F1-score. The experimental results show that the Random Forest model achieves superior performance and is therefore selected for further explainability analysis. To improve the interpretability of the prediction model, three explainability methods Random Forest Feature Importance, SHAP, and Permutation Importance are used. The normalized scores from these methods are combined to introduce the Mental Health Influence Score (MHIS), offering a consistent and dependable ranking of factors that impact children’s mental health. The analysis finds that average daily screen time is the most impactful factor, with the educational-to - recreational ratio, age, and primary device usage ranking next underscoring the critical role of behavioral factors in mental health evaluation. The proposed MHIS framework enhances the transparency and reliability of feature importance analysis by integrating several explainability techniques rather than relying on a one method. Furthermore, the integration of LLM-based interpretation improves the practical usability of the framework by converting numerical importance scores into human-readable explanations. This enables researchers, educators, parents, and healthcare professionals to better understand the relationship between screen-time behaviours and mental health outcomes, thereby supporting more informed decision-making. In future work, the proposed framework can be extended by incorporating larger and more diverse datasets, advanced deep-learning models, and real-time mental health monitoring systems. Additional behavioural, environmental, and psychological factors may also be combined to further improve prediction performance and interpretability. Moreover, the proposed MHIS framework can be validated across different age groups and populations to evaluate its generalizability and applicability in broader mental health assessment scenarios.

1.    F. Karim, A. Oyewande, L. F. Abdalla, R. C. Ehsanullah, and S. Khan, “Social Media Use and Its Connection to Mental Health: A Systematic Review,” Cureus, vol. 12, no. 6, Art. no. e8627, 2020, doi: 10.7759/cureus.8627. 2.    X. Yang and G. Li, “Psychological and Behavioural Insights From Social Media Users: Natural Language Processing–Based Study,” JMIR Formative Research, 2025.  3.   J. M. Twenge and W. K. Campbell, “Associations Between Screen Time and Lower Psychological Well-Being Among Children and Adolescents: Evidence From a Population-Based Study,” Preventive Medicine Reports, vol. 12, pp. 271–283, 2018, doi: 10.1016/j.pmedr.2018.10.003. 4.    J. Gao et al., “Mental Health Problems and Social Media Exposure During COVID-19 Outbreak,” PLOS ONE, vol. 15, no. 4, Art. no. e0231924, 2020, doi: 10.1371/journal.pone.0231924. 5.    H. C. Woods and H. Scott, “: Social Media Use in Adolescence Is Associated With Poor Sleep Quality, Anxiety, Depression and Low Self-Esteem,” Journal of Adolescence, vol. 51, pp. 41–49, 2016.  6.  Pachava, O. A. Lasekan, S. K. Golla, and S. Gosikonda, “Machine Learning Analysis of Social Media's Impact on Mental Health of Indian Youth,” International Research Journal of Multidisciplinary Scope, vol. 5, no. 2, pp. 623–635, 2024, doi: 10.47857/irjms.2024.v05i02.0592.  7.  K. Shahzadi, M. S. Sarfraz, M. Saqib, and H. M. Shah, “Machine Learning Methods for Analysing the Impact of Social Media on Students' Academic Performance,” Journal of Applied Artificial Intelligence, vol. 1, no. 1, pp. 13–22, 2020, doi: 10.48185/jaai.v1i1.96. 8.     Y. Xu, L. Xie, M. Bao, X. Yang, S. Chen, and Z. Gao, “Exploring Subjective Well-Being in Adolescents: The Role of Mental Health and Addictive Behaviours Through Machine Learning,” International Journal of Mental Health Promotion, vol. 27, no. 5, pp. 667–682, 2025, doi: 10.32604/ijmhp.2025.062808.  9.  S. M. Lundberg and S. I. Lee, “A Unified Approach to Interpreting Model Predictions,” in Advances in Neural Information A. 10. Barredo Arrieta et al., “Explainable Artificial Intelligence (XAI): Concepts, Taxonomies, Opportunities and Challenges Toward Responsible AI,” Information Fusion, vol. 58, pp. 82–115, 2020, doi: 10.1016/j.inffus.2019.12.012. Processing Systems, vol. 30, 2017.