MEIT: Memory Efficient Itemset Tree for Targeted Association Rule Mining

The Itemset Tree is an efficient data structure for performing targeted queries for itemset mining and association rule mining. It is incrementally updatable by inserting new transactions and it provides efficient querying and updating algorithms. However, an important limitation of the IT structure, concerning scalability, is that it consumes a large amount of memory. In this paper, we address this limitation by proposing an improved data structure named MEIT (Memory Efficient Itemset Tree). It offers an efficient node compression mechanism for reducing IT node size. It also performs on-the-fly node decompression for restoring compressed information when needed. An experimental study with datasets commonly used in the data mining literature representing various types of data shows that MEIT are up to 60 % smaller than IT (43% on