PMML Research And Data Mining System Platform Design

The vast applications of RDBMS impels the emergence of Data Mining technology. Data Mining System will have been used in finance, telecommunication and retail industry. After a mass of data mining system had been used, there is an urgent need to develop a universal data mining language, just like SQL in the RDMBS, between different data mining system manufacturers, between different research institutions and between every data mining system manufacturer and every research institution. Still there some nonstandard languages used by some companies and institutions, for instance, data mining query language(DBMiner's DMQL),data mining modeling language(OMG's PMML) and general data mining language (Microsoft's OLE DB for Data Mining). In this article, we introduce the concepts and functions of data mining, discusse and analyses the development of data mining system and its interrelated technologies, introduce and analyse PMML. Based on the research in software architecture, in this article, we design a data mining system platform, which can provide a experiment platform for data mining technology researcher. New data mining algorithms can easily be integrated into this platform if they comply with the data model interface and mining model interface of this platform. This platform is a data mining system framework, practical, tightly combined with data mining applications,and easily updated and transplanted. The platform integrates lately PMML standard and component technology into data mining system, thus, it has nice expansibility. The platform mainly consisted of data server, mining server and expert server. Data server includes data abstract, the representation of meta data and data access interface. Mining sever include mining functions. It creates a data mining algorithm type architecture, using TminingAlgorithm class as it parent class, throught mining functions abstraction.