New method for mining frequent itemsets with between-item positive correlation

Low support makes dramatic increase in the number of itemsets and brings less efficient frequent itemset mining. Correlation measures introduced to restrict the number of frequent itemsets generated in order to improve the efficiency of mining under certain conditions. An improved FP-Tree algorithm using node linked list FP-Tree is proposed. This algorithm exploits efficient pruning strategies using a between-item positive correlated differences measure with a good antimonotone. Non-positive correlated long model and invalid itemsets are filtered. The range of support threshold allowing mining is expanded. Experimental results indicate the given algorithm is efficient and feasible.