Pattern discovery in probabilistic databases

Abstract

Modeling probabilistic data is one of important issues in databases due to the fact that data is often uncertainty in real-world applications. So, it is necessary to identify potentially useful patterns in probabilistic databases. Because probabilistic data in 1NF relations is redundant, previous mining techniques don’t work well on probabilistic databases. For this reason, this paper proposes a new model for mining probabilistic databases. A partition is thus developed for preprocessing probabilistic data in a probabilistic databases. We evaluated the proposed technique, and the experimental results demonstrate that our approach is effective and efficient.