Federated Machine Learning Using The Internet Of Medical Things For Cardiac Disease Detection

Abstract

The proliferation of medical data sourced from healthcare organizations has led to an increased utilization of Machine Learning (ML) methodologies within the medical domain. In order to uphold the effectiveness of healthcare professionals, it is imperative to design machine learning models that are dependable and credible. Machine learning (ML) models are widely employed in the field of healthcare to generate accurate disease predictions, while simultaneously ensuring the protection of data gathered via Internet of Medical Things (IoMT) devices. Federated Learning (FL) techniques are well-suited for the preservation of Internet of Medical Things (IoMT) data due to their ability to preserve only the trained models and advance with information from distributed users. Fluorescence techniques possess the potential to significantly revolutionize the medical business by enabling rapid disease diagnosis, hence improving the effectiveness of therapy. The efficacy of these FL approaches is diminished as a result of the substantial volume of data being transmitted between local and remote locations. To address this concern, a unique methodology called FedEDFA is introduced, which integrates Federated Machine Learning with a meta-heuristic optimization algorithm. The utilization of the Enhanced Dragonfly Optimization algorithm is employed to effectively choose the pertinent features and thereafter utilize them for disease prediction. This strategy enhances the system's resilience in networking situations that are prone to insecurity. The efficacy of the proposed FedEDFA is evaluated by its application to the UCI Cleveland dataset, with the objective of predicting cardiovascular illness while ensuring data security and privacy. The proposed approach demonstrates a greater level of accuracy, measuring at 98.3%, when compared to other existing methods.