Health Sense: Real-Time AI-IoT Health Risk Prediction

Abstract

With the rapid growth of Internet of Things (IoT) technologies,real-timehealthmonitoringhasbecomeincreasingly efficient and accessible. This paper presents HealthSense, an intelligent healthcare system that integrates IoT-based wearable sensors with machine learning techniques to predict potential health risks in real time. The system continuously collects physiological parameters—heart rate, oxygen saturation (SpO2), bodytemperature,andbloodpressure—andprocessesthismulti- modal data using an ensemble pipeline of Random Forest, Support Vector Machine (SVM), and Long Short-Term Memory (LSTM) networks. These models enable early anomaly detection anddelivertimelyalertsforpreventiveclinicalintervention. A user-friendly web dashboard provides real-time visualisation, trendanalysis,andpersonalisedhealthinsights.Experimen- tal evaluation on labelled physiological datasets demonstrates that the proposed LSTM-based model achieves a classification accuracy of 94.2% with an average prediction latency belowthreeseconds,outperformingconventionalbaselines.HealthSense enhances preventive healthcare by enabling proactive, data- driven decision-making and reducing the burden on emergency clinical services.

Introduction

The text describes HealthSense, an AI-powered IoT-based remote patient monitoring system designed to continuously track vital health parameters and detect early signs of medical emergencies in patients with chronic diseases.

The motivation comes from the rising global burden of chronic illnesses like cardiovascular disease, diabetes, and respiratory conditions, where traditional hospital-based or periodic checkups fail to capture early warning signs. HealthSense addresses this by enabling continuous, real-time physiological monitoring using wearable sensors.

The system integrates a multi-layer architecture consisting of:

• An IoT sensing layer with wearable devices measuring heart rate, SpO2, ECG, temperature, and blood pressure
• A cloud communication layer using MQTT and HTTP for fast data transmission
• A preprocessing pipeline to clean noisy sensor data, handle missing values, normalize signals, and extract features like heart rate variability
• A hybrid machine learning layer combining Random Forest, SVM, and LSTM models
• A dashboard layer that provides visualization and sends tiered alerts to clinicians

The ML system uses a stacked ensemble approach, where outputs from all models are combined to improve prediction accuracy and reliability. The LSTM helps capture time-based physiological patterns, while Random Forest and SVM handle structured feature relationships. Risk is classified into normal, moderate, and high-risk categories, triggering alerts accordingly.

The system is trained and evaluated using datasets like MIMIC-III and UCI Heart Disease, and it achieves strong performance, with the stacked ensemble reaching 94.2% accuracy and an AUC of 0.97.

Experimental results show that the system can process data and generate alerts in under 2 seconds end-to-end, making it suitable for real-time healthcare monitoring. Case studies demonstrate that it can detect conditions like arrhythmias and oxygen desaturation minutes earlier than conventional monitoring systems.

Conclusion

HealthSense represents a comprehensive and validated ad- vancement in real-time AI-IoT healthcare by unifying multi- parameter wearable sensing, cloud-based preprocessing, and a stackedLSTM-basedmachinelearningensembleintoasingle, deployableplatform.Thesystemachieves94.2%classification accuracy,anAUC-ROCof0.97,andsub-2-secondend-to-end prediction latency, outperforming all evaluated baselines. By shifting healthcare from a reactive to a proactive paradigm, HealthSense has the potential to reduce preventable hospitali- sations, lower healthcare costs, and improve patient quality of life.Theopenresearchdirectionsoutlined—edgeinference, federated learning, and clinical validation—chart a clear path toward clinical deployment and large-scale population health monitoring.

References

[1] A.KumarandR.Sharma,“IoT-basedhealthmonitoringandalertsystemusing machine learning,” IEEE Access, vol. 9, pp. 103915–103928,2021. [2] P. Verma, S. Sood, and S. Kalra, “Cloud-centric IoT based studenthealthcare monitoring framework,” Journal of Ambient Intelligence andHumanized Computing, vol. 9, no. 5, pp. 1–17, 2018. [3] T. Nguyen, A. W. Tran, and M. S. Islam, “Deep learning for predictivehealthcare monitoring: A survey,” IEEE Reviews in Biomedical Engi-neering, vol. 16, pp. 78–99, 2023. [4] World Health Organization, “Noncommunicable diseases: Key facts,”Sep. 2022. [Online]. Available: https://www.who.int/news-room/fact-sheets/detail/noncommunicable-diseases [5] E. J. Topol, “High-performance medicine: The convergence of humanand artificial intelligence,” Nature Medicine, vol. 25, no. 1, pp. 44–56,2019. [6] Z. Liang, G. Ward, and C. Wen, “Motion artefact removal from photo-plethysmographic signals,” Biomedical Signal Processing and Control,vol. 60, p. 101947, 2020. [7] M. Cvach, “Monitor alarm fatigue: An integrative review,” BiomedicalInstrumentation & Technology, vol. 46, no. 4, pp. 268–277, 2012. [8] P.Rajpurkaretal.,“Cardiologist-levelarrhythmiadetectionwithconvo-lutional neural networks,” arXiv:1707.01836, 2017. [9] S. M. R. Islam et al., “The internet of things for health care: Acomprehensive survey,” IEEE Access, vol. 3, pp. 678–708, 2015. [10] Q. Mao et al., “Multicentre validation of a sepsis prediction algorithm,”BMJ Open, vol. 8, no. 1, 2018. [11] A. E. W. Johnson et al., “MIMIC-III, a freely accessible critical caredatabase,” Scientific Data, vol. 3, p. 160035, 2016. [12] Z. C. Lipton et al., “Learning to diagnose with LSTM recurrent neuralnetworks,” arXiv:1511.03677, 2015. [13] Z. Che et al., “Recurrent neural networks for multivariate time serieswith missing values,” Scientific Reports, vol. 8, p. 6085, 2018. [14] S. N. Shukla and B. M. Marlin, “Interpolation-prediction networks,”ICLR, 2019. [15] W.Shietal.,“Edgecomputing:Visionandchallenges,”IEEEIoTJournal, vol. 3, no. 5, pp. 637–646, 2016. [16] N.Riekeetal.,“Thefutureofdigitalhealthwithfederatedlearning,”npj Digital Medicine, vol. 3, 2020. [17] K.Heetal.,“Deepresiduallearningforimagerecognition,”CVPR,2019. [18] S.Tulietal.,“HealthFog:Smarthealthcaresystemusingdeeplearning,”Future Generation Computer Systems, vol. 104, pp. 187–200, 2020. [19] H.B.McMahanetal.,“Communication-efficientlearningofdeepnetworks,” AISTATS, 2017.

Copyright

Copyright © 2026 Tasneem Banu, Tejaswini N, Sirisha T, Sunitha V. 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.