The Component Design And Development Of Data Mining Platform

Component Based Development provides developers with a method for building applications that are flexible and able to accommodate the ever-changing demands of users in a cost-efficient and timely manner. In this article, the author presents the design of a DataMining platform using component-based approach. This platform provides an experimental platform for datamining researchers, complying with our data object interface and mining model interface, new mining algorithms can be easily integrated to our system. Our platform also is a valuable system skeleton for datamining application. Complied with PMML 2.0 Specification, using component-oriented technology makes the system an open and easy-to-upgrade solution.The article begins with an introduction of the conception and method of data mining. After that, it briefly introduces and analyseses the development of datamining system software and some important technologies that are bound up with datamining system, then PMML is introduced and analysesed. The author then put due emphasis on the architecture design, function design, component design, and implementation technologies of the system.Based on J2EE Server-Side-Architecture, system design is mainly made up of three J2EE Server: DataServer.MiningServer and ExpertServer. DataServer represents the data abstraction, it includes meta-data reprensentation and source data access. MiningServer implements the mining functions. An algorithm class hierarchy is established to implement different type of mining functions. The result has been expressed as PMML object and is the base data for ExpertServer to realize its predicating function. PMML-compliant format makes the ExpertSystem capable of using mining model derived from different mining tools.