An Efficient Algorithm for Incremental and Interactive High Utility Itemset Mining

High utility itemset mining (HUIM) is an important data-mining task. Most of existing algorithms for HUIM do not consider transaction addition and deletion. When a database is updated, they need to scan the whole database to rebuild their data structures. To deal with this problem, an efficient tree structure IHUP-Tree is proposed. IHUP-Tree can be adjusted efficiently when a transaction is added into or deleted from a database. Incremental HUIM can be performed efficiently based on IHUP-Tree. IHUP-Tree can also be applied in interactive HUIM. The algorithm based on IHUP-Tree discovers high utility itemsets (HUIs) in two phases. In phase I, an over-estimated technique is adopted to set an upper bound for the utility of an itemset in the database. The itemsets whose over-estimated utilities are no less than a user-specified minimum utility threshold are selected as candidates. In phase II, the candidates are verified by scanning the database one more time. However the algorithm based on IHUP-Tree generates too many candidates, and it is time-consuming to verify them. Thus in this paper we proposed a novel tree structure IHUh-Tree and an efficient algorithm IHUI-Miner for incremental and interactive HUIM. Different from the algorithm based on IHUP-Tree, IHUI-Miner does not generate any candidate. Extensive performance analyses show our proposed tree structure is efficient, and our algorithm is at least one order of magnitude faster than the state-of-the-art algorithm in increment and interactive HUIM.