Carlos Eduardo Barbosa,  Eduardo Ogasawara,  Daniel de Oliveira,  Marta Mattoso. 

Classical data mining (DM) process is usually composed by a chain of programs and data that are modeled by a specialist. This chain of programs may be modeled as a workflow and takes advantage of scientific workflow management systems (SWfMS). This paper proposes an approach for parallelizing tasks within a data mining workflow. Experimental results using time­series forecast with neural networks reinforce the performance gains and additional benefits provided by SWfMS, such as provenance recording of the workflow.

http://www.lbd.dcc.ufmg.br:8080/colecoes/waamd/2009/013.pdf

Processando consulta.