FIMHAUI: Fast Incremental Mining of High Average-Utility Itemsets
İrfan Yıldırım, Mete Çelik · 2018
The IMHAUI (Incremental mining of high average-utility itemsets) algorithm is presented to find high average-utility itemsets (HAUIs) efficiently from incremental database. It uses a compact tree structure named IHAUI-Tree (Incremental high average utility itemset tree) to maintain the information of all accumulated transactions in the incremental environment. Then, a pattern growth approach is employed to generate a set of essential candidate itemsets using the tree. However, it needs to recursively extract conditional pattern bases and builds conditional local trees in the process of mining candidate itemsets, so this will increase the time required for discovering the set of candidate itemsets. Therefore, this paper proposes a tree data structure, named mIHAUI-Tree (modified IHAUI-Tree) and an efficient algorithm, called FIMHAUI (Fast incremental mining of high average-utility itemsets). FIMHAUI uses the mIHAUI-Tree structure to store required information of transactions. When a mining request is occured, FIMHAUI first adjusts the mIHAUI-Tree to maximize node sharing effect of the mIHAUI-Tree. Then, it extracts the projected database from mIHAUI-Tree. Finally, it uses database projection and transaction merging techniques to discover candidate itemsets, efficiently. Experimental results show that FIMHAUI outperforms the existing IMHAUI algorithm in terms of the runtime.