Secure and Weighted Hybrid Machine Learning Framework for Accurate Academic Stress Monitoring

Abstract

Stress among college students has increasingly become a serious concern, affecting both academic performance and overall well-being. This study presents the development and implementation of a privacy- aware, hybrid machine learning system designed to monitor and manage student stress levels in a college environment. An Android application was developed to collect real-time data on students’ physiological and behavioural indicators, including Heart Rate, Sleep Hours, Academic Load, Screen Time, Physical Activity, and Mood Score. The collected data were analysed using a Hybrid Random Forest–SVM model, where feature importance weighting and adaptive probability fusion were applied to enhance predictive accuracy. Additionally, AES encryption was integrated to ensure the secure handling of sensitive student information. Based on the assessed stress levels, the system provides tailored guidance and supportive strategies aimed at reducing psychological strain and promoting mental wellness. The proposed approach SWH Framework allows continuous monitoring and early identification of high stress levels, enabling timely intervention. Experimental results demonstrate that the system achieves 92% accuracy, outperforming existing single-model approaches. These findings suggest that hybrid, privacy-aware mobile health technologies can serve as effective, reliable, and accessible tools for supporting stress management among college students and enhancing their overall academic experience.

Introduction

Psychological stress among college students has become a major public health concern due to academic pressure, social challenges, financial burdens, and career uncertainty. Prolonged stress can negatively affect academic performance, mental health, and overall well-being, potentially leading to anxiety, depression, and reduced productivity. Early detection and effective management are therefore essential in academic environments.

Recent advancements in the Internet of Things (IoT) and the Internet of Medical Things (IoMT) have enabled real-time health monitoring through smart and wearable devices. When combined with machine learning (ML), these technologies enhance predictive accuracy and automated decision-making in healthcare systems. Although existing stress detection systems often rely on wearable sensors and deep learning models, most focus on clinical environments or physiological signals alone. Limited research addresses mobile-based stress monitoring systems tailored specifically for college students within academic settings.

Proposed Solution

To address this gap, the study proposes a Secure Weighted Hybrid (SWH) stress monitoring framework, implemented as an Android-based mobile application. The system enables real-time stress monitoring, early identification of elevated stress levels, and timely intervention strategies. It provides a scalable, low-cost, and practical solution for academic institutions.

Key Features of the Proposed System:

• Hybrid Random Forest–Support Vector Machine (RF-SVM) model

• Feature importance weighting based on Random Forest

• Adaptive probability fusion to improve classification reliability

• AES-256 encryption for data privacy and security

• Real-time monitoring using mobile-collected behavioral and self-reported data

• Improved predictive accuracy of 92%, outperforming traditional single-model approaches (80–87%)

Research Gap Identified

1. Most systems rely on single ML models with moderate accuracy.

2. Many approaches depend solely on wearable sensors without integrating multi-feature academic data.

3. Feature importance weighting is rarely applied.

4. Data privacy is often neglected.

5. Few systems are designed specifically for academic stress monitoring.

Objectives

The study aims to:

• Develop a hybrid RF-SVM model for accurate stress prediction.

• Apply feature importance weighting to emphasize key indicators (heart rate, sleep hours, academic load, mood score).

• Use adaptive probability fusion to reduce misclassification.

• Ensure data confidentiality using AES encryption.

• Evaluate performance using Accuracy, Precision, Recall, and F1-score.

• Provide a deployable real-time stress monitoring solution for colleges.

Methodology

Data Collection

Data were collected from 1000 students via a mobile application. Features included:

• Sleep Hours

• Heart Rate

• Screen Time

• Physical Activity

• Academic Load

• Mood Score

Stress level was categorized as binary (Low = 0, High = 1).

Data Processing

• Missing values replaced using mean imputation.

• Standardization applied for scale normalization.

• Dataset split: 70% training, 30% testing (stratified sampling).

Security

• Data encrypted using AES-256 (CBC mode) during storage and transmission.

• Decryption performed securely during model training.

Hybrid Model (SWH Framework)

1. Random Forest (RF) calculates feature importance.

2. Features are weighted using normalized importance scores.

3. Support Vector Machine (SVM) classifies stress levels using weighted features.

4. Adaptive Probability Fusion combines RF and SVM outputs to generate final predictions.

This combination improves robustness and reduces misclassification.

Results

The proposed hybrid model achieved:

Key Observations:

• 91% of low-stress cases correctly classified.

• 93% of high-stress cases correctly identified.

• Balanced performance across both classes.

• Significant improvement over baseline (~0.83 accuracy before optimization).

Performance improvements were achieved through:

• Larger dataset size

• Feature weighting

• Hyperparameter tuning

• Adaptive probability fusion

Conclusion

This study presents a privacy-aware hybrid RF-SVM based SWH Framework for real-time stress prediction among college students. By combining Random Forest feature importance weighting with SVM classification and adaptive probability fusion, the proposed model effectively captures critical stress indicators such as Heart Rate, Sleep Hours, Academic Load, and Mood Score. The integration of AES encryption ensures the secure handling of sensitive student data, addressing a key limitation in existing stress monitoring systems. Experimental results demonstrate that the proposed system achieves 92% accuracy, outperforming previous models that relied on single classifiers or wearable sensors. Overall, this framework provides a robust, reliable, and privacy- preserving solution for academic stress monitoring, offering potential for real-time deployment in educational environments. Future work can explore expanding the dataset, integrating wearable IoT sensors, and implementing federated learning to further enhance predictive accuracy and scalability.

References

[1] Kadhim, K.T.; Alsahlany, A.M.; Wadi, S.M.; Kadhum, H.T. An Overview of Patient’s Health Status Monitoring System Based on Internet of Things (IoT). Wirel. Pers. Commun. 2020, 114, 2235–2262. [Google Scholar] [CrossRef] [2] Mohammed, C.M.; Askar, S. Machine learning for IoT healthcare applications: A review. Int. J. Sci. Bus.2021, 5, 42–51. [Google Scholar] [3] Sadad, T.; Safran, M.; Khan, I.; Alfarhood, S.; Khan, R.; Ashraf, I. Efficient Classification of ECG Images Using a Lightweight CNN with Attention Module and IoT. Sensors 2023, 23, 7697. [Google Scholar] [CrossRef] [4] Talaat, F.M.; El-Balka, R.M. Stress monitoring using wearable sensors: IoT techniques in medical field. Neural Comput. Appl. 2023, 35, 18571–18584. [Google Scholar] [CrossRef] [PubMed] [5] Jalali, E.; Soltanizadeh, H.; Chen, Y.; Xie, Y.M.; Sareh, P. Selective hinge removal strategy for architecting hierarchical auxetic metamaterials. Commun. Mater. 2022, 3, 97. [Google Scholar] [CrossRef] [6] Chen, Y.; Xu, R.; Lu, C.; Liu, K.; Feng, J.; Sareh, P. Multistability of the hexagonal origami hypar based on group theory and symmetry breaking. Int. J. Mech. Sci. 2023, 247, 108196. [Google Scholar] [CrossRef] [7] He, Z.; Shi, K.; Li, J.; Chao, J. Self-assembly of DNA origami for nanofabrication, biosensing, drug delivery, and computational storage. iScience 2023, 26, 106638. [Google Scholar] [CrossRef] [8] Jayadev, P.G.; Bellary, S. A hybrid approach for classification and identification of iris damaged levels of alcohol drinkers. J. King Saud. Univ-Comput. Inf. Sci. 2022, 34, 5273–5285. [Google Scholar] [9] Rao, Z.; Tung, P.Y.; Xie, R.; Wei, Y.; Zhang, H.; Ferrari, A.; Klaver, T.P.; Körmann, F.; Sukumar, P.T.; Kwiatkowski da Silva, A.; et al. Machine learning–enabled high-entropy alloy discovery. Science 2022, 378, 78–85. [Google Scholar] [CrossRef] [10] Sun, W.; Guo, Z.; Yang, Z.; Wu, Y.; Lan, W.; Liao, Y.; Wu, X.; Liu, Y. A review of recent advances in vital signals monitoring of sports and health via flexible wearable sensors. Sensors 2022, 22, 7784. [11] Rachakonda, L.; Mohanty, S.P.; Kougianos, E.; Sundaravadivel, P. Stress-Lysis: A DNN-Integrated EdgeDevice for Stress Level Detection in the IoMT. IEEE Trans. Consum. Electron. 2019, 65, 474–483. [Google Scholar] [CrossRef] [12] Khan, A.R.; Saba, T.; Sadad, T.; Nobanee, H.; Bahaj, S.A. Identification of anomalies in mammograms through internet of medical things (IoMT) diagnosis system. Comput. Intell. Neurosci. 2022, 2022,1100775. [Google Scholar] [CrossRef] [PubMed] [13] Rosenzweig, C.; Karoly, D.; Vicarelli, M.; Neofotis, P.; Wu, Q.; Casassa, G.; Menzel, A.; Root, T.L.Estrella, N.; Seguin, B.; et al. Attributing physical and biological impacts to anthropogenic climate change.Nature 2008, 453, 353–357. [Google Scholar] [CrossRef] [PubMed] [14] Alanazi, H.O.; Abdullah, A.H.; Qureshi, K.N. A critical review for developing accurate and dynamic predictive models using machine learning methods in medicine and health care. J. Med. Syst. 2017, 41,69. [Google Scholar] [CrossRef] [PubMed] [15] Atzori, L. The internet of things: A survey. Comput. Netw. 2010, 54, 2787–2805. [Google Scholar] [16] Nath, R.K.; Thapliyal, H.; Caban-Holt, A.; Mohanty, S.P. Machine Learning based Solutions for Real-Time Stress Monitoring. IEEE Consum. Electron. Mag. 2020, 9, 34–41. [Google Scholar] [CrossRef] [17] Lin, S.; Zhang, H.; Gao, Y.; Du, M.; Vai, M. The Effects of Muscle Stress on Signal Transmission in the Intra-Body Communication. In Proceedings of the 2016 IEEE International Conference on Consumer Electronics-China (ICCE-China), Guangzhou, China, 19–21 December 2016; pp. 1–3. [Google Scholar] [18] Magaa, V.C.; Muoz-Organero, M. Reducing Stress on Habitual Journeys. In Proceedings of the 2015 IEEE 5th International Conference on Consumer Electronics - Berlin (ICCE-Berlin), Berlin, Germany, 6–9 September 2015; pp. 153–157. [Google Scholar] [19] Ciabattoni, L.; Ferracuti, F.; Longhi, S.; Pepa, L.; Romeo, L.; Verdini, F. Real-Time Mental Stress Detection based on Smartwatch. In Proceedings of the 2017 IEEE International Conference on Consumer Electronics (ICCE), Las Vegas, NV, USA, 8–10 January 2017; pp. 110–111. [Google Scholar] [20] Lawanot, W.; Inoue, M.; Yokemura, T.; Mongkolnam, P.; Nukoolkit, C. Daily Stress and Mood Recognition System Using Deep Learning and Fuzzy Clustering for Promoting Better Well-Being. In Proceedings of the 2019 IEEE International Conference on Consumer Electronics (ICCE), Las Vegas, NV, USA, 11–13 January 2019; pp. 1–6. [Google Scholar] [21] Nath, R.K.; Thapliyal, H.; Caban-Holt, A. Validating Physiological Stress Detection Model Using Cortisol as Stress Bio Marker. In Proceedings of the 2020 IEEE International Conference on Consumer Electronics (ICCE), Las Vegas, NV, USA, 4–6 January 2020; pp. 1–5. [Google Scholar] [22] Lee, J.-M.; Byun, W.; Keill, A.; Dinkel, D.; Seo, Y. Comparison of Wearable Trackers Ability to Estimate Sleep. Int. J. Environ. Res. Public Health 2018, 15, 1265. [Google Scholar] [CrossRef] [23] Arnold, J.A.; Cheng, Y.; Baiani, Y.; Russell, A.M. Systems and Techniques for Tracking Sleep Consistency and Sleep Goals. US Patent 20 170 347 946A1, 2 June 2016. [Google Scholar] [24] Karydis, A.M. Methods, Systems, and Apparatus for Self-Calibrating EEG Neurofeedback. US Patent 20 160 235 324A1, 15 February 2016. [Google Scholar] [25] Sannholm, F.; Paalasmaa, J.; Leppakorpi, L. System for Determining the Quality of Sleep. US Patent 20 160 213 309A1, 31 December 2015. [Google Scholar] [26] Bone, D.; Lee, C.; Chaspari, T.; Gibson, J.; Narayanan, S. Signal processing and machine learning for mental health research and clinical applications. IEEE Signal Process. Mag. 2017, 34, 196–195.

Copyright

Copyright © 2026 A. Janita, B. NagaLakshmi, R. Ramachandiran . This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.