Application of Federated Learning Approaches for Time-Series Classification in eHealth Domain

Contemporary Machine Learning approaches (e.g., Deep Learning) need huge volumes of data to build accurate and robust statistical models. Nowadays, such data are often collected by distinct and geographically distributed entities and successively transmitted to and stored by centralized nodes that implement the learning process. However, this approach exposes data to security and privacy issues during the transmitting process. In order to demonstrate the benefits which Federate learning can bring concerning these issues, this paper proposes an application of the Federated Learning approach to training a time series (TS)-based model for the early identification of the level of risk associated with patients with hypertension in a federated healthcare environment. The results demonstrate the goodness of the solution in comparison with traditional centralized approaches. The solution achieves good results; indeed, the accuracy is over 0.90%. The goodness of the model is also confirmed by the values of recall and precision, equal to 82% and 96%, respectively.