Knowledge Discovery Based On Agent

In this paper, we present the definition of Agent, the deal with action's process and inside mechanism, in detail. At the theory of Agent, this paper present a state change based action process, the action of Agent take place when state of Agent changed, so it can make the Agent system more steady, and easy to find error when the system work in a wrong way. Then, we research of the multi-Agent system (MAS). In this section, we present a multi-Agent model "agora", it came out being though simulating the agora. It is a more flexible multi-Agent model, and has exactly universality, more it can make communication of each Agent in multi-Agent system more convenient. In this paper we focus on the structure of agora, the different from traditional blackboard is that all Agent that in this structure are administered by the supervisor in the agora, and it can match the information of the Agent in the agora, so accomplish cooperation of the Agents.In the rear part of the paper, we combine the agora model with the process of knowledge discovery, and present Agent-based knowledge discovery technique. The technique focus on frequent pattern mining, the existed models of frequent pattern mining have the advances respectively, but there is shortcoming in the intelligence and universality. There are some differences in the mining technique of the item set, sequence etc, the Agent-based knowledge discovery technique can mine the item set. sequence and tree with one technique, so it can improve the intelligence and universality of the knowledge discovery system in a great extent.