The Distributed Web Log Mining System And Its Implementation Based On Multi-Agent Technology

Traditional DM (Data Mining), basically, is a data-analyzing tool for local data at present and only can produce few of generalized or understood knowledge from local datasets. With the accumulating and distributing of data store, better methods or tools for dealing with these increasing distributed data are needed greatly. Research of Multi-Agent technology has become a hotspot in the domain of Distributed Artificial Intelligence, and propels the progress of many research areas since 1990s.This research aims at applying Multi-Agent technology to distributed data mining. This paper includes three parts. The first part discusses the basic theory of Multi-Agent technology and distributed data mining, at the end of this part, paper introduces distributed Web log mining technology. At the second part, paper introduced a distributed Web log mining system based on Multi-Agent technology, which is divided into User-Access layer, Agent layer and Server layer. Agent layer is composed of User Agent (UA), Mining Agent (MA) and Resource Agent (RA), and a Remote Adapter at Server layer is designed to finish the job of data preprocessing. This system structure is a new try in the area of distributed Web log mining, which can be extended and consummated in future work. It has a great deal of potential both in research and application. The third part of this paper designed and incremented each component of this system structure. In this part of research, author studied the communication mechanism of agent and its implementation method, designed and developed a Structural Communicating Body (SCB), defined its interface, then the primitive set of the system is designed combining the characteristics of SCB. More over, pager brings forward the design method of the Remote Adapter and Mining Agent of the system, based on the research of Web log mining technology.The distributed Web log mining system based on Multi-Agent technology discussed in this paper is feasible, however, since the work is a new try in the area ofdistributed Web log mining, there were few related materials. So, although a lot of work has been spent studying each core technology, it is still in the state of theory researching at present. After further researching and perfection, it will surely bring forth more value of theory and application both in Multi-Agent technology area and Distributed Web log mining area.