Speed-up iterative frequent itemset mining with constraint changes

Mining of frequent itemsets is a fundamental data mining task. Past research has proposed many efficient algorithms for this purpose. Recent work also highlighted the importance of using constraints to focus the mining process to mine only those relevant itemsets. In practice, data mining is often an interactive and iterative process. The user typically changes constraints and runs the mining algorithm many times before being satisfied with the final results. This interactive process is very time consuming. Existing mining algorithms are unable to take advantage of this iterative process to use previous mining results to speed up the current mining process. This results in an enormous waste of time and computation. In this paper, we propose an efficient technique to utilize previous mining results to improve the efficiency of current mining when constraints are changed. We first introduce the concept of tree boundary to summarize useful information available from previous mining. We then show that the tree boundary provides an effective and efficient framework for the new mining. The proposed technique has been implemented in the context of two existing frequent itemset mining algorithms, FP-tree and tree projection. Experiment results on both synthetic and real-life datasets show that the proposed approach achieves a dramatic saving of computation.