Discovering general partial orders in event streams

Sequence of time-ordered events arise in a variety of applications like customer transaction databases, alarm sequences in telecommunication networks, fault logs in manufacturing plant data, web interaction logs, etc. A popular framework for temporal pattern extraction from such data is the frequent episode discovery paradigm. An episode is a set of nodes with a partial order prescribed on it, with each node associated with an event type. Efficient algorithms exist for episode discovery when the associated partial order is total(serial episode) or trivial(parallel episode). In this paper, we propose efficient algorithms for discovering frequent episodes with general partial orders. The algorithms generalize the existing apriori-based discovery algorithms for serial and parallel episodes. There is an inherent combinatorial explosion in frequent partial order mining. We point out that frequency alone is not a sufficient measure of interestingness for general episodes. We present post-processing techniques to prune an explosive number of uninteresting patterns from the set of frequent partial orders . We demonstrate with simulation the efficiency of our algorithms and the effectiveness of the post-processing filters. The simulations show utility of partial order mining in unearthing information that cannot be discovered by serial/parallel