UC Irvine

Many algorithms have been proposed to solve the association rule learning problem. However, most of them suffer from the problem of scalability either because of unacceptable time complexity or tremendous memory usage, especially when the dataset is enormous and the minimum support (minsup) is low. This paper introduces a new approach that follows the divide-and-conquer paradigm, which can exponentially reduce both the time complexity and memory usage, even on a single machine.