Research On The Application-independent Databases Clustering Technology

A great company has many subsidiaries distributed in different zones, and the greatcompany utilizes multi-database system to organize its transactions. Knowledge discoveryin multi-database is a kind of pressing need for the organizers of the great company.Transfering the data of all subsidiaris to the head company simply, then assembly miningthese crude data may produce many defects. So, developing a kind of multi-databasemining system is a new challenge in data mining domain. Application-dependent miningtechnology through merging databases after choosed is low efficient and easy to lose someuseful information; distributed mining technology dose not produce middle-rules and ishard to accomplish parrallel mining algorithm. Application-independent mining technologybasing multi-database’ best-group renews the process of mining multi-database focusing onthe three phases of the normal data mining process. In the phase of preparing data, the newjob about multi-database’ best-group will be done; in the second phase, mining the localdatabase of every subsidiay seperately, and gaining local association-rule set; and in thephase of expressing and evaluating knowledge, the job about analysing and synthesizingrules will be done: analysing rules will produce high-voting rules, exceptional rules,suggesting rules, and synthesizing rules will produce not only grobal rules but also grouprules. The gorup-rule owns both higher degree of support and higher degree of confidencethan the corresponding global-rule. The mission of mining multi-database has been finishedintegrally and completely by the new data mining process including three phases.The main research job of this paper includes three aspects, as below:1. abstracting the basic information of the research of the multi-database miningtechnology. It contains the importance, the basic structure and characteristics of theproblem of mining multi-database; the quality, defects of the two kinds of existingmulti-database mining technology; the computing method of the similarity between twodatabases, the grouping method of databases; and the databases classification technology inmulti-database mining.2. presenting a kind of best-group technology basing fuzzy clustering analysis. This kindof technology has evident advantages comparing to the existing two kinds of multi-databasemining technology, and its integer time consumption is very low, it can get the moreeffective information supporting the globe decision. And more, the design of thistechnology’s core algorithm is a kind of innovation, which makes fuzzy analysis of thesymbol data record possible.3. presenting the concept of group-rule, and utilizing the synthesis technology on theassociate rule to confirm the advantage of them. The group-rule is globle rule in a groupspecially, which limit the decision scope through assembling the objects for decision, andcut down the consumption, but enhances the effectness of the decision.