Mining top-k distinguishing periodic patterns in temporal databases.

Distinguishing pattern mining, which is an important task in data mining, aims to find the differences between collections of data. However, the existing distinguishing pattern mining methods fail to consider the problem that some patterns appear regularly in one class of data sets and irregularly in another class, which may lose some interesting patterns. To fill this gap, we propose a new problem named mining top-k distinguishing periodic patterns in temporal databases. We aim to find k patterns with the most significant difference between two classes of data sets with a period constraint. And we introduce an algorithm called kDPP-Miner to solve this problem. Several pruning rules are designed to improve its efficiency. The experimental results in real data sets demonstrate that our algorithm is effective and efficient.