关键词: 频繁项目集, 并行挖掘, FP-Growth, Map/Reduce

Abstract: Algorithm FP-Growth is a classic algorithm for mining frequent item sets which is based on frequent pattern tree. In order to improve the efficiency of algorithm FP-Growth for mining association rules from massive datasets, parallel FP-Growth algorithm FPPM is presented. The algorithm is based on Map/Reduce model, and the local frequent pattern tree of each computing node is built, these local trees are mined to get local frequent item sets, and local frequent item sets are merged into global frequent item sets. After the statistics of the local frequent item sets, a complete result is got. In this paper, the idea of FPPM is introduced and its performance is studied. The experimental results show that the parallel algorithm FPPM has high scalability.

Key words: frequent item set, parallel mining, FP-Growth, Map/Reduce