Study On Data Mining Technologies Based On Mobile Agent

With the development of computer technology, network technology and Internet technology, the technology of data mining has developed rapidly. This technology is engaged in identifying the mode which is valid, novel, potentially useful and ultimately comprehensible from large, incomplete, noisy, fuzzy and stochastic data sets, and achieves finding decision-making support of mode. In practice, most data sets distribute geographically in several locations, therefore, if the data is proposed to be collected in one place (such as data warehouse), high-speed communication network will be required, as a result, time of response will be longer and security will be destroyed. However, technology of distributed data mining resolves the above problems well and in distributed data mining system, client, data, hardware and software needed by data mining can be distributed geographically.In order to improve the communication efficiency and platform compatibility of distributed data mining and solve the security problems existing in the fields of data transmission etc., the paper researched on how to apply the technology of mobile agent in distributed data mining to build distributed data mining system which is of remote interaction, platform independency, bandwidth saving, higher maintainability and flexibility. Based on"Diagnosis of Semi-health State of Network Engineering System and Studies on Intelligent Maintenance Decision-making Mechanism"granted by Chongqing Jiaotong University Graduate Education Innovation Fund (No.0904), the paper researched and analyzed the application of mobile agent in distributed data mining in depth. Main achievements of the paper include:(1) Technology of mobile agent was comprehended and grasped in depth, and characters and system structure etc. of mobile agent were deeply analyzed here. The merits and drawbacks of other distributed computing models were analyzed, and on this basis, a distributed data mining model based on mobile agent was proposed. Furthermore, on the base of analyzing the factors which influence mining capability, cost models of several typical distributed data mining models were researched. System framework of Aglet platform (development platform of mobile agent), life cycle & basic action of Aglet and communication mechanism between Aglets were studied here.(2) Making use of Aglets of mobile agent platform, a prototype of distributed data mining system was designed and realized, in which the characters of mobile agent such as mobility, autonomy and reactivity were utilized, the controlling station dispatched Aglets to remote stations to carry out mining task, and returned partial mining results to the controlling station. In the system, only a small quantity of code, partial mining results and state of Aglets needed to move, and consequently, communication traffic were decreased.(3) On foundation of study on theory of associative rules mining and algorithms of distributed associative rules mining, an improved algorithm of FDM was put forward: algorithm of DDM_AR_BA. In this algorithm, because of using mobile agent, communication mode among stations in distributed condition was thoroughly changed, and through theoretical analysis, partial mining results of remote stations could be obtained after Aglets moving out and home twice between the controlling station and target station (the Aglets which moved out and home twice were different Aglets bearing different mining tasks.), that could greatly decrease communication cost and increase mining efficiency. The experiment on data in UCI data sets testified the validity of the algorithm, and superiority of mining efficiency of this algorithm would be more evident especially for tremendous amounts of data.