Research Into Agent-based Data Mining Model And Application To Lava Fracturing Solution

With the high-speed development of database and information technology, more and more data accumulated in the various areas, so it's a big challenge to dig out the useful information from a lot of data. Data mining is a process to find out the interesting knowledge from vast data. The traditional data mining technology behaved not so well at intelligence, it emphasizes on artificial participates actively, which led to the complication of data mining process and the low efficiency in the circular mining process. The profound research of agent technology and its application in the data mining break a new path for the solution of above-mentioned problems.With Agent technology and data mining technology, this article advanced a three-layer data mining model that based on a multi-agent corporate system, solved the problems of model's organization structure, the design of each agent, the corporation of each agent, task analysis and task schedule, etc. This model and the circumstance could interact, and its integration of data mining algorithm demonstrated its intelligence, initiative and openness. This model adopted a three-layer (user, management and the actualization of tasks) corporate work mode. In the layer of task actualization, each algorithm agent could work parallel, which enhanced the executive efficiency and the precision of data mining. In the process of task analysis, the management agent would confirm the best analytical project with the use of key route, which optimized the work capability of system and reduced the consumption of system resources.Combined to the demand analysis of lava fracturing operation according to the above model, we designed an assistant analysis system of lava fracturing operation based on agent. This article explained the structure and function of user agent and management agent. The system finished the data mining task of lava fracturing operation with decision tree algorithm and association rule algorithm, and put out a part of test results in the end of the article.