Domain-Oriented Data Mining System Study And Implement

Along with the growth and development of enterprises, The information in the business enterprise is increasingly complicated. Especially in the process of establishing data mining system, a lot of problems arise such as difficulties of communication and collaboration between different areas, frequent changes in the needs of corporations and information model, poor system to be expanded , low rate of development efficiency and reuse.With the combination of domain engineering, domain framework, object-oriented technology, software component technology, data warehouse and data mining, this paper presents a development model of domain-oriented data mining system and a new data mining hierarchical structure which is similar to OSI network model. Furthermore we design a framework of domain-oriented data mining system.In order to support flexible construction of data warehouse, uniform standard layer was added between data source of tradition and data warehouse model. In order to guarantee the transparency access and the scalability model, two mapping methods were adopted between uniform view and resource data, between data warehouse and uniform view. Based on this architecture, construction method of uniform view model and mapping algorithms amongst data warehouse three-layer were given to complete ETL function effectively.This paper reduces the complexity of implementation of data mining system by layered strategy. The system is divided into four layers: scheme layer, task layer, algorithm layer, data warehouse layer. The result is that the mining scheme is separated from the problem of data mining and analysis, the mining task is separated from the mining scheme, the mining algorithm is separated from mining task, the data which need to be dealt is separated from mining task. So different mining problems can be the same scheme, different schemes can have the same task, different tasks can have the same algorithm, different tasks can have the same data.At last, this paper briefly introduces several data mining models of the system and further works which need to do.