AI-Driven Predictive Analytics for Early Disease Detection Leveraging Body Sensor Networks and Advanced Machine Learning Models

Abstract

Early disease detection is very essential in the field of healthcare as a means of intervention early and improving patient outcomes. The paper provides an evaluation of different AI predictive analytics models to use in body sensor networks. The models that the authors have assessed include Deep Learning Ensemble, Attention Mechanism LSTM, Variational Autoencoder, Recurrent Neural Network (RNN), Capsule Networks, Neural Architecture Search, and Federated Learning Model. The evaluation metrics used to assess the models were accuracy, precision, recall, and AUC-ROC. The physiological data for this study consists of signals such as heart rate, body temperature, body acceleration and other physical activity levels recorded by several body sensor networks (BSNs) over time. With over 2000 participants, we ensure that the dataset is rich enough and covers a vast range of demographics and health conditions to provide a good foundation for model generalization. Findings indicate that the Deep Learning Ensemble model performed the best with an accuracy of 95.1 indicating that sophisticated machine learning algorithms can produce high accuracy early disease detection.