Mining uncertain data with probabilistic guarantees

TLDR

Data uncertainty is inherent in many applications, and probabilistic databases have been developed to manage it, but the exponential number of possible worlds makes this technically challenging. This work investigates frequent pattern and association rule discovery from probabilistic data under Possible World Semantics, proposing two efficient bottom‑up and top‑down algorithms. The algorithms efficiently enumerate frequent patterns, can be extended to maximal patterns, and the resulting patterns can be used to generate association rules. Experiments on real and synthetic datasets confirm the methods’ effectiveness.

Abstract

Data uncertainty is inherent in applications such as sensor monitoring systems, location-based services, and biological databases. To manage this vast amount of imprecise information, probabilistic databases have been recently developed. In this paper, we study the discovery of frequent patterns and association rules from probabilistic data under the Possible World Semantics. This is technically challenging, since a probabilistic database can have an exponential number of possible worlds. We propose two effcient algorithms, which discover frequent patterns in bottom-up and top-down manners. Both algorithms can be easily extended to discover maximal frequent patterns. We also explain how to use these patterns to generate association rules. Extensive experiments, using real and synthetic datasets, were conducted to validate the performance of our methods.