Association Rules Discovery via Approximate Method from Probabilistic Database

Abstract

Association rules and frequent patterns discovery is always a hot topic in database communities. As real data is often affected by noise, in this paper, we study to find frequent patterns and generate association rules over probabilistic database under the Possible World Semantics. This is technically challenging, since a probabilistic database can have an exponential number of possible worlds. Although several efficient algorithms are proposed in the literature, there is still a large space for improvement due to the redundancy property of frequent patterns over probabilistic data. To address this issue, we employ approximate idea and propose a more efficient algorithm to mine frequent pattern. After that, we present two distinct strategies to obtain the association rules, and design an evaluation model to measure the accuracy of association rules result. Finally, extensive experiments have been done on real databases, demonstrating that the proposed method preforms better than state-of-art methods in most cases.