Comparative Analysis of Federated Association Rules in a Simulated Environment for Medical Applications

Abstract

This article examines some of the most relevant algorithms for association rule mining in a medical context, within the framework of unsupervised Federated Learning (FL) in a simulated environment. Unlike traditional algorithms that rely on centralized databases, FL operates on decentralized devices, prioritizing data privacy and algorithm efficiency. This emerging paradigm in machine learning and data mining aims to preserve privacy in edge networks. In this study, the privacy of subject data is ensured as the algorithms operate and collaborate between nodes, sharing only the results of their computations while keeping raw data encapsulated and encrypted. The work evaluates key aspects such as execution time and encryption/decryption efficiency in edge networks. This analysis is motivated by the increasing demand for data analysis in the healthcare sector, where maintaining data privacy is critical due to the proliferation of data privacy regulations.