Chaining Patterns

Finding condensed representations for pattern collections has been an active research topic in data mining recently and several representations have been proposed. In this paper we introduce chain partitions of partially ordered pattern collections as high-level condensed representations that can be applied to a wide variety of pattern collections including most known condensed representations and databases. We analyze the goodness of the approach, study the computational challenges and algorithms for finding the optimal chain partitions, and show empirically that this approach can simplify the pattern collections significantly.