Smart IoT Health Monitoring System with AI-based Anomaly Detection

Continuous health monitoring is vital for vulnerable populations but remains largely inaccessible. As the demand for remote, proactive healthcare grows, the Internet of Medical Things (IoMT) shows great promise in bridging this gap. This paper aims to develop a new, patient-centric Edge-to-Cloud system that uses effective machine learning to predict a patient's real-time clinical risk based on continuous vital signs. Sensors based on an affordable ESP32 microcontroller measure critical indications like heart rate, blood oxygen (SpO2), and body temperature, seamlessly transmitting data to the ThingSpeak cloud. The true novelty lies in pairing this hardware with a cloud-based "brain" engineered to overcome class-imbalance challenges. To find the proverbial needle in the haystack, we evaluated Support Vector Machine (SVM) and Random Forest classifiers. Random Forest emerged as the superior model, achieving an 88.13% accuracy and a 0.878 F1-score. Crucially, it minimized false alarms— reducing them by more than half compared to the SVM model—ensuring that real, lifethreatening anomalies are prioritized over routine noise. This powerful synergy turns a basic sensor into a proactive guardian, providing highly reliable real-time anomaly detection and empowering caregivers with a continuous safety net that catches emergencies the instant they happen.