PADMA-RAD 2025: A Modernized Multi-Agent System for Distributed Association Rules Mining

Abstract

This paper presents PADMA-RAD 2025, an evolution of the classic PADMA system that integrates a multi-agent system (MAS) architecture with the Intelligent Data Distribution (IDD) algorithm for distributed association rule mining. While preserving the core PADMA structure and the IDD algorithm, we adapt them to modern constraints such as intermittent connectivity, local processing on embedded devices, and integrated security. The proposed architecture is evaluated through a prototype implementation. Experimental results on standard datasets demonstrate a significant performance improvement, with a reduction in execution time from 31.7s for classical data mining to 5.1s for our PADMA+IDD approach, alongside an increase in perceived precision. A comparative analysis with existing systems, including HealthAgents, confirms the relevance and competitiveness of our approach for real-world applications like connected health monitoring in resource-constrained environments.