Efficient Incremental Itemset Tree for approximate Frequent Itemset mining on Data Stream
Pavitra Bai S., Ravi Kumar G K · 2016
Mining frequent itemsets and association rules on data stream is an important and challenging task. Tree based approaches have been extensively studied and widely used for their parallel processing capability. Itemset Tree is an efficient data structure to represent the transactions for performing selective mining of frequent itemsets and association rules. The transactions are inserted incrementally and provide on-demand ad-hoc querying on the tree for fining frequent itemsets and association rules for different support and confidence values. However the size of tree grows larger for unbounded data streams limiting the scalability. In this paper, we propose an Approximation based Incremental Memory Efficient Itemset Tree (AIMEIT) algorithm which is an extension to Memory Efficient Itemset Tree (MEIT) algorithm to construct the itemset tree from data stream. The user defined minimum support of interest has been used along with Lossy counting algorithm to prune transactions before inserting them into the tree. Experimental results show that the proposed algorithm is more memory efficient and takes lesser processing time for constructing the tree.