Improved algorithm for mining N-most interesting itemsets

Abstract

BOMO algorithm constructs conditional FP-Tree recursively so that it requires more memory and CPU resources. To solve this problem, an algorithm for mining N-most interesting itemsets based on COFI-Tree is presented. This algorithm adopts COFI-Tree. COFI-Tree doesn't need to construct conditional FP-Tree recursively and there is only one COFI-Tree in memory at a time. Experiment shows that (1) the new algorithm based on COFI-tree performs faster than current best algorithm BOMO;(2) the algorithm has good performance for large data set, especially it shows the best when for k value is smaller than 4. © 2011 Springer-Verlag.