@inproceedings{f527d494c59c439cbc3465721cd93230,
title = "Advanced Stress Classification and Vital Signs Forecasting for IoT-Health Monitoring",
abstract = "As wearable sensors and IoT technologies evolve, the demand for real-time health monitoring systems increases, particularly in stress detection and vital signs forecasting. This paper presents an ensemble-based approach for stress classification using CatBoost, LightGBM, AdaBoost, and HGBM, supported by explainability tools SHAP. It also introduces a hybrid forecasting system for oxygen saturation (SpO2) and pulse using GRU, LSTM, ARIMA, and AR. The system is validated on realworld datasets and deployed on ESP32 with a MAX30102 sensor for real-time use. Results show that classification accuracies of 97.59\% and 97\% can be achieved for Non-EEG dataset and for WESAD dataset, respectively. It is also demonstrated that 0.33 and 0.25 MAE for pulse and SpO2 are attained, respectively.",
keywords = "Deep Learning, Ensemble Learning, Forecasting, Health Monitoring, IoT, Stress Detection, Vital Signs",
author = "Abubakar Danasabe and Hossain, \{Md Moazzem\} and Jahiduzzaman, \{F. M.\} and Rabie, \{Khaled M.\}",
note = "Publisher Copyright: {\textcopyright} 2025 IEEE.; IEEE International Conference on E-health Networking, Applications and Services, IEEE HealthCom 2025 ; Conference date: 21-10-2025 Through 23-10-2025",
year = "2025",
doi = "10.1109/HealthCom60686.2025.11342957",
language = "English",
series = "2025 IEEE International Conference on E-health Networking, Application and Services, Healthcom 2025",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
booktitle = "2025 IEEE International Conference on E-health Networking, Application and Services, Healthcom 2025",
address = "United States",
}