A Scalable Strategy for Mining Association Rules under Grids

Sequential Association Rule Mining (ARM) algorithms are characterized by a high computational complexity due to two facts: (i) they have to mine a very large search space (ii)they have high demands of database access. Association rule mining technique have progressively been adapted to large-scale systems in order to benefit from the large-scale computing capabilities and the huge storage capacity provided by theses systems. Performance issues (i.e, efficiency and scalability) are determinant factors for association rule mining algorithms [1]. In this paper we present an important part of our multilevel strategy that aims to improve the scalability of distributed ARM algorithms. Our main goal is to obtain a running time that grow linearly in proportion with the size of the database, given the available system resources (i.e., available computing nodes, their main memory and their disk space, etc.). The French research grid "Grid’5000" is used as our experimental test-bed.