Research On Grid Data Mining Technology

As the next generation of network, grid has strong computation power, excellent system expansibility and efficient ability of managing distributed resource. It breaks the limit of computation power, storage, resource distributing and the way of sharing resource, so it is adapted to be used in computation intensive and data intensive applications.Grid data mining technology research is the hot topic of recent years. It is the combination of data mining and grid technology which take full advantage of all kinds of services provided by grid platform to implement distributed data mining. The services such as task management, task scheduling and resource management provided by grid computation are benefit to distributed data mining.Firstly, in this paper, the representative grid data mining technology——Knowledge Grid and GridMiner are analyzed deeply, and their shortages are pointed out too. By analyzing the current grid architecture and comparing the advantage and disadvantage of OGSI(Open Grid Service Infrastructure) and WSRF(Web Service Resource Framework), this paper chooses GT4(Globus Toolkit 4), which supports WSRF, as the basic platform of grid data mining architecture.Besides, to resolve varieties of data mining tasks which users submit, this paper proposes the data mining task decomposing policy based on application fields and tasks in sub field. The use of ontology to describe the data mining task can ensure users execute grid data mining pellucidly.Then, according to the feature of distributed data mining, an innovative distributed grid data mining architecture based on GT4 is proposed. It adopts ontology, data integration, service discovery and composition as the core technology. The main functions of all modules and the ontology base are designed. The relationship of all modules and the workflow of grid data mining on this architecture are also discussed in details. Finally, an instance of data mining service deployment on grid is presented.