Efficient incremental mining of contrast patterns in changing data

A contrast pattern, also known as an emerging pattern [7], is an itemset whose frequency differs significantly between two classes of data. Such patterns describe differences between datasets and have been shown to be useful for building powerful classifiers [11, 9, 2, 8] . Incrementally mining them in changing data is very important, where transactions can be inserted and deleted and mining needs to be repeated after changes occur. When the changes are small, the previously mined contrast patterns should be reused where possible, to compute the new patterns. A primary example of changing data is a data stream a sequence of continuously arriving transactions (or itemsets). Mining of contrast patterns in a data stream is useful for stream classification [2] and network traffic change detection [4]. Work in [10] presented an algorithm to incrementally mine contrast patterns, but is oriented to updates of a single type. When a dataset changes due to insertion and deletion together, the efficiency of [10]’s approach is reduced, due to redundant computations. In this paper, we present a new algorithm that addresses the scenario of incrementally mining contrast patterns in response to simultaneous insertion and deletion. Our ap-