Smartbaseg : uma arquitetura baseada em conhecimento e em separação de interesses para a melhoria da qualidade de aplicações de mineração de dados no grid

AUTOR(ES)
DATA DE PUBLICAÇÃO

2006

RESUMO

Although the Grid seems to be a very promising platform for supporting the Data Mining (DM) process, a lot of challenges still need to be overcome in order to attain an ideal integration between these two domains. In this dissertation, a software architecture, called SMARTBASEG, is defined with the purpose of improving the quality of data mining applications on the Grid. More specifically, SMARTBASEG looks for obtaining more efficiency, maintainability and portability for those applications. One of SMARTBASEG s suppositions is that knowledge about DM, Grid and performance optimization heuristics can be represented and, afterwards, exploited during the DM applications executions on the Grid for improving the efficiency of those executions. This declarative feature of the architecture enables its optimization layer to decide, dynamically, how to transform DM procedures into tasks and jobs, aiming at an effective scheduling and load balancing. The other supposition that this work is based on is that using the separation of concerns principle enables DM applications developed for the Grid to achieve the maintainability and portability level of serial versions of those applications. To do this, SMARTBASEG makes available a set of DM abstractions by means of components and templates so that the developer can keep his/her attention on DM domain, thus avoiding him/her to worry about Grid concerns and/or performance issues. By this way, the DM applications development and the source code of those applications are not depreciated, since the Grid access is encapsulated and becomes transparent to the developer.

ASSUNTO(S)

informÁtica - dissertaÇÕes sistemas de informacao

ACESSO AO ARTIGO

Documentos Relacionados