Classical Quantum Federated Learning for Mental Stress Detection

Abstract

The growing prevalence of mental stress has created a strong need for intelligent systems capable of early and accurate detection. This work proposes a hybrid learning framework that combines classical machine learning, quantum computing techniques, and federated learning to classify stress levels in in-dividuals. Initially, physiological signals from the WESAD dataset are processed to remove noise and normalize feature values. A Random Forest model is then applied to identify the most influential features, which helps reduce data dimensionality and makes the model suitable for quantum processing. The selected features are transformed into quantum states using rotation-based angle encoding with RY gates. A variational quantum neural network is designed to learn patterns within the encoded data through parameterized quantum circuits. To ensure data privacy and enable decentralized training, a federated learning strategy is incorporated, where multiple clients train models locally and share only model parameters. These parameters are combined using a federated averaging method to form a global model without exposing sensitive data. The developed system categorizes mental stress into three levels: low, medium, and high. The results indicate that integrating quantum learning with federated approaches can effectively handle sensitive health data while maintaining reliable classification performance. This study demonstrates the potential of combining emerging computational paradigms to build secure and efficient healthcare prediction systems.

Introduction

This paper presents an intelligent mental stress detection framework that combines Classical Machine Learning (CML), Quantum Machine Learning (QML), and Federated Learning (FL) to accurately classify stress levels while preserving user privacy. As mental stress has become a major public health concern due to increasing academic, professional, and lifestyle pressures, the proposed system addresses the limitations of traditional stress assessment methods, which rely on subjective self-reporting or clinical evaluations. By utilizing wearable sensor data and advanced computational techniques, the framework enables continuous, automated, and secure stress monitoring.

The proposed system uses physiological signals from the WESAD dataset, including electrodermal activity (EDA), heart rate (ECG), respiration, and other biometric measurements. After preprocessing and normalization, a Random Forest algorithm performs feature selection to identify the most informative features and reduce data dimensionality. These selected features are then encoded into quantum states using rotation-based (RY gate) encoding, enabling processing by a Variational Quantum Neural Network (VQNN), which classifies stress into three categories: low, medium, and high.

To protect sensitive healthcare data, the framework incorporates Federated Learning, where multiple clients train local models using their own data without sharing raw information. Instead, only model parameters are transmitted to a central server, which combines them using the Federated Averaging (FedAvg) algorithm to produce a global model. This decentralized approach enhances privacy while allowing collaborative learning across distributed data sources.

The literature review highlights recent advancements in stress detection using federated learning, quantum computing, explainable AI, and Internet of Medical Things (IoMT) technologies. Although previous studies have demonstrated improvements in classification accuracy and privacy preservation, they often focus on individual techniques rather than integrating classical machine learning, quantum neural networks, and federated learning into a unified framework. The proposed work addresses this research gap by combining all three approaches to improve both prediction performance and data security.

Experimental evaluation compared the performance of three models. The Classical Extra Trees classifier achieved the highest accuracy of 98.4%, correctly classifying nearly all stress samples with minimal errors. The Quantum Machine Learning model achieved 91.4% accuracy, demonstrating stable performance and highlighting the potential of quantum computing for future healthcare applications. The Federated Learning model achieved 86.3% accuracy, providing secure, privacy-preserving distributed learning, although performance was slightly reduced due to decentralized training and variations in local data distributions. Across all models, the medium-stress class experienced the highest misclassification rate because of dataset imbalance.

References

[1] Gupta, M. Kumar Maurya, K. Dhere, V. Kumar Chaurasiya, “Privacy-Preserving Hybrid Feder ated Learning Framework for Mental Health-care Applications: Clustered and Quantum Approaches,” IEEE Access, 2024. [2] Chao Ren, Rudai Yan, Huihui Zhu, Han Yu, Yan Xu, Minrui Xu, “Toward Quantum Federated Learning,” 2025 [3] ] Shiyi Jiang, Yuan Li, Farshad Firouzi, Krishnendu Chakrabarty, “Federated Clustered Multi-Domain Learning for Health Monitoring,” 2024 [4] Abdulrahman Alhamadi, Haroon Khan, Ghufran Shafiq, Junaid Ahmed, Bakhtiar Ali, Muham mad Javed, “A Privacy-Preserved IoMT-Based Mental Stress Detection Framework with Federated Learning,” 2023 [5] Q. Wu, X. Chen, Z. Zhou, J. Zhang, “FedHome: Cloud-Edge Based Personalized Federated Learning for In-Home Health Monitoring,” IEEE Transactions on Mobile Computing, 2022. [6] Shoaib Ullah, Madam Hussain Shah, Adeel Anjum, “Quantum-Enhanced Federated Learning with Differential Privacy,” 2024. Avail-able: https://arxiv.org/pdf/2301.08361.pdf [7] V. Meena, G. Raghavender, S. Jayakar, J. Senthil Kumar, “Privacy-Focused Federated Learning for Mental Health Data with XAI Tech-niques,” ICDICI Conference Proceedings, 2025 [8] H. Lee, J. Park, S. Kim, “Quantum Federated Learn-ing for Medical Image Classification,” 2023. Available: https://arxiv.org/pdf/2301.08361.pdf [9] A. Reiss, D. Stricker, “Federated Learning for Wearable Health Moni-toring,” 2022. Available: https://arxiv.org/pdf/2203.06735.pdf [10] S. Y. C. Chen, S. Yoo, “Federated Quantum Machine Learning,” 2021 [11] A. Sierra-Sosa et al., “Quantum Classifiers for Biomedical Data,” 2022 [12] L. Cao et al., “DeepMood: A Mobile Mood Detection Framework,” 2017. [13] Mahdi Chehimi, Samuel Yen-Chi Chen, Walid Saad, Don Towsley, Me´rouane Debbah, “Foundations of Quantum Federated Learning Over Classical and Quantum Networks,” 2023. [14] Nouhaila Innan, Muhammad Al-Zafar Khan, Alberto Marchisio, Muhammad Shafique, Mohamed Bennai, “FedQNN: Federated Learning using Quantum Neural Networks,” 2024. [15] Amandeep Singh Bhatia, David E. Bernal Neira, “Federated Hierarchical Tensor Networks: A Collaborative Learning Quantum AI-Driven Frame-work for Healthcare,” 2024. [16] Rafael Ballester et al., “Quantum Federated Learning: A Comprehensive Literature Review,” 2025. [17] Dinh C. Nguyen, Md Raihan Uddin, Shaba Shaon, Ratun Rahman, Octavia Dobre, Dusit Niyato, “Quantum Federated Learning: A Com-prehensive Survey,” 2025. [18] M. Balasubramani et al., “A Novel Framework for Quantum-Enhanced Federated Learning for Healthcare Applications,” Sensors, 2025. [19] D. Godavarthi et al., “Federated Quantum-Inspired Anomaly Detection using Distributed Learning,” Frontiers in Artificial Intelligence, 2025. [20] Micah J. Sheller et al., “Multi-Institutional Deep Learning Modeling Without Sharing Patient Data: A Feasibility Study on Brain Tumor Segmentation,” 2020.

Copyright

Copyright © 2026 Aarya Lambore, Prajwal Thorat, Sarthak Wayase, Vaibhav Jadhav, Mr. Digambar Padulkar. 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.