ABSTRACT
Internet of things (IoT)-based systems generate vast and continuous streams of health data that require robust handling for accurate and timely health interventions. Concept drift, a common phenomenon in data streams that adopt statistical properties that change over time, is a significant challenge. This chapter describes an adaptive machine learning approach and dynamic adaptive weighted ensemble (DAWE) that are designed to effectively handle concept drift and improve prediction accuracy in real-time health monitoring. The novel value of DAWE is the use of the exponential weighted moving average (EWMA) technique, which adjusts model weights based on prediction errors while continuously refining the model’s overall predictive performance. Our model performance shows a superior accuracy of 96.5%, significantly outperforming several conventional models. These findings emphasize the importance and efficacy of employing adaptive learning techniques to manage the challenges posed by high-speed, high-volume data streams in IoT-based health monitoring systems.
