Improvement and realization of association rules mining algorithm based on FP-tree

Traditional FP-growth algorithm adopts FP-tree structure to express association of item sets in transaction sets and finds all of frequent item sets recursively. The algorithm increases the time complexity and the space complexity in calculating conditional pattern base, because it backtracks the same paths many times. As to the above defects, a FPIFM algorithm is presented in the paper. The algorithm stores all of precursor nodes of every node in the node domain, then the sub-condition pattern base of every node are calculated. Finally, sub-condition pattern base are combined and ergodic nodes are released. Experimental result shows that FPIFM algorithm is superior to the traditional FP-growth algorithm.