A New Fast Frequent Itemsets Mining Algorithm Based on Forest

Frequent itemsets mining is an important data mining task. Until now there are many algorithms to mine frequent itemsets, this paper emphatically analyses existing algorithms´ memory structure of transaction database and realization skill, as well as the usage of memory. On the basis, two new data structures UFP-Tree and FP-Forest are designed, which use multi-trees structure to store data and solve the single FP-tree storage large scale database difficulty problem. At the same time, a fast frequent itemsets mining algorithm, F-Fminer is presented, which adopts divide and conquer strategy to mine frequent itemsets for every UFP-Tree with deepth-first searching method, then performs right shift combination operation for the branches in UFP-Tree. In order to reduce the time spending of allocation and deallocation memory, this paper also designed a high performance memory manager. According to experimentation on real data sets, the algorithm has greatly enhanced frequent itemsets mining efficiency.