Mining strongly closed itemsets from data streams

We consider the problem of mining strongly closed itemsets from transactional data streams. Compactness and stability against changes in the input are two characteristic features of this kind of itemsets that make them appealing for different applications. Utilizing their algebraic and algorithmic properties, we propose an algorithm based on reservoir sampling for approximating this type of itemsets in the landmark streaming setting, prove its correctness, and show empirically that it yields a considerable speed-up over a straightforward naive algorithm without any significant loss in precision and recall. As a motivating application, we experimentally demonstrate the suitability of strongly closed itemsets to concept drift detection in transactional data streams.