Machine Learning–Based Prediction and Diagnosis of Arrhythmia Using Multimodal Data

Wearable health monitoring systems have rapidly transformed healthcare by enabling continuous, non-invasive, real-time monitoring of physiological parameters. This study presents a predictive health analytics framework that applies machine learning to multimodal data collected from wearable devices such as fitness bands and smartwatches. Embedded IoT sensors capture vital signs including body temperature, SpO₂ and pulse rate, which then undergo signal pre-processing and model-driven prediction. To detect health anomalies and anticipate potential risks, supervised learning algorithms, namely Decision Trees, Random Forests and Support Vector Machines, are compared. The proposed model supports an integrated proactive alerting mechanism, scalability for large user bases and edge processing. Publicly available multivariate physiological recordings from PhysioNet are used to train and evaluate the models, which supports clinical relevance and reproducibility. Early findings suggest that the framework can improve the early detection of medical problems and support hospital-centred monitoring. The work contributes to intelligent, widely accessible healthcare solutions and lays a foundation for future personalised health prediction systems.