Mining Optimized Support Rules for Numeric Attributes

In this paper, we generalize the optimized support association rule problem by permitting rules to contain disjunctions over uninstantiated numeric attributes. For rules containing a single numeric attribute, we present a dynamic programming algorithm for computing optimized association rules. Furthermore, we propose a bucketing technique for reducing the input size, and a divide and conquer strategy that improves the performance significantly without sacrificing optimality. Our experimental results for a single numeric attribute indicate that our bucketing and divide and conquer enhancements are very effective in reducing the execution times and memory requirements of our dynamic programming algorithm. Furthermore, they show that our algorithms scale up almost linearly with the attribute's domain size as well as the number of disjunctions. 1 Introduction Association rules, introduced in [AIS93], provide a useful mechanism for discovering correlations among the underlying data. In it..