基于项头表节点的Fp-growth改进算法

An improved FP-growth algorithm based on item head table node

关联规则中的Fp-growth算法是不产生候选集的代表,将原算法FP-tree和项头表的Node_link字段删除,把Ln当作项头表。对任意频繁项ai,首先找到所有FP-tree节点的item-name与ai的项名相同的节点,对每个树节点寻找它的频繁模式,找到频繁项ai的所有频繁模式可节省1/5树的空间,把Ln当作项头表,省去项头表的空间,从而提高算法效率。实验结果表明,改进后的算法性能优于原算法性能。

In the association rule＇ s mining, FP-growth algorithm is one of the most high effective algorithm, and it doesn＇ t produce the candidate sets. It deletes the original FP-tree algorithm and the node_link field in the item header table, makes Ln as the item header table. For any frequent item ai, at first, it finds all the FP-tree nodes whose item-name has the same item name with ai, looking for the frequent pattern for every node, it can save 1/5 tree space when we found the frequent pattern for every frequent item, makes Ln as the item header table, saves head table space in order to improve the efficiency of the algorithm. The experimental results show that the improved algorithm is better than the original algorithm in performance.