The Research On Key Techniques Of Data Mining System Based On Web Services And The Design Of The Prototype

With the rapid increasement of data, the enterprises want to mine the knowledge behind the large amounts of data in order to support decision. Some existing data mining tools such as Intelligence Miner, Enterprise Miner supply rich data mining functions, but these tools can't mine distributed and heterogeneous data on Internet/Extranet, and they can't integrate with operating system effectively and have no pertinences. If the enterprises use these tools, they will expand more money but many mining functions become useless and it's very difficult to upgrade the algorithms library. This thesis puts forward a data mining system architecture based on Web services. It can integrate with existing operating system and can mine the data in distributed database. Because of using the Web Services, it can be independent of platform and programming language, ease to be deployed and to manage algorithms library flexibly.Firstly, this thesis gives a general architecture about data mining system based on Web Services and introduces Web Services technique which realizes Services-Oriented Architecture. And then it researches the key techniques of data mining system in detail, including data preprocess, algorithm management, building algorithm library and data mining model visualization. Using PMML to express the result model of data mining, it can share and reuse the data mining model. And then it researches association rules mining algorithm and implement it. At last it gives the implementation of data mining system prototype based on Web Services using B/S architecture, and has mined association rules in the market basket data of one supermarket. The result proves that the architecture of data mining system is expansible, believable and feasible.