From parallel data mining to grid-enabled distributed knowledge discovery

Abstract

Data mining often is a compute intensive and time requiring process. For this reason, several data mining systems have been implemented on parallel computing platforms to achieve high performance in the analysis of large data sets. Moreover, when large data repfositories are coupled with geographical distribution of data, users and systems, more sophisticated technologies are needed to implement high-performance distributed KDD systems. Recently computational Grids emerged as privileged platforms for distributed computing and a growing number of Grid-based KDD systems have been designed. In this paper we first outline different ways to exploit parallelism in the main data mining techniques and algorithms, then we discuss Grid-based KDD systems. © Springer-Verlag Berlin Heidelberg 2007.