AgentDiscover: A Multi-Agent System for Knowledge Discovery from Databases

AgentDiscover is a multi-agent based intelligent recommendation system, supporting real-time control and feedback, for building and execution workflows for knowledge discovery from databases (KDD). The aim of the proposed system is to deal with the complexity of KDD processes and to offer a tool that supports both researchers exploring KDD methods and non-expert users looking for quick results in this field. A prototype was developed in JADE and the approach was tested for a medical dataset.