Study On Multi Agent-based Enterprise Distributed Association Rules Discovery

It becomes an key issue for up-to-date enterprises that how to mine requisite knowledge from distributed databases. In consideration of data security, privacy and consistency, as well as real-time and accuracy of knowledge, distributed knowledge discovery system is more adapted to enterprises than data warehouse. However, Some issues have been found in present distributed knowledge discovery model, such as only local and global knowledge can be mined except region knowledge which is very important for enterprises, data mining has been operated directly on transaction database and this will impact the throughput of OLTP and only some models without data mining algorithm, task scheduling algorithm and knowledge integration algorithm have been proposed. It's hard to use under practical condition.In order to solve these issues, an enterprise distributed association rules discovery model based on multi-Agent has been presented in this thesis. In the model, management agent schedules all decomposed sub data mining tasks according to SA algorithm. Execute agent executes data mining tasks on remote sites with task details and data mining algorithm-DMFIF. Knowledge agent collects local knowledge from result agent of all local sites, and gains the region knowledge or global knowledge by integrating all local knowledge according to KI algorithm. Finally, a prototype system for the distributed knowledge discovery system has been carried out.