Faster association rules for multiple relations

Abstract

Several algorithms have already been implemented which combine association rules with first order logic formulas. Although this resulted in several usable algorithms, little attention was payed until recently to the efficiency of these algorithms. In this paper we present some new ideas to turn one important intermediate step in the process of discovering such rules, i.e. the discovery of frequent item sets, more efficient. Using an implementation that we coined FARMER, we show that indeed a speed-up is obtained and that, using these ideas, the performance is much more comparable to original association rule algorithms. 1