Date of Publication :29th March 2018

Abstract: Many algorithms and techniques are developed for enumerating itemsets from transactional databases. They produce a large number of frequent itemsets when the minimum support threshold is low and /or the dataset is dense. A large number of discovered patterns makes further analysis of generated patterns troublesome. Therefore, it is important to find a small number of representative patterns to best approximate all other patterns. This paper modifies the algorithm MinRPset for finding representative pattern sets. It follows the same concepts such as - covered and greedy method in MinRPset but, it uses NCFP-tree instead of CFP-tree for storing frequent itemsets in a compressed manner. The experiment results show that our algorithm gives better execution time to generate representative patterns sets for the mushroom dataset in an efficient manner comparing to the algorithm MinRPset

Reference :