A Parallel Association Rule Mining Algorithm Based on Spark and its application

The research of association rule mining (ARM) is very hot. Along with the increase of data size, single machine ARM algorithms have encountered difficulties in time and memory. Distributed computing could break this bottleneck. In this paper, a parallel association rule mining algorithm is proposed, which is called PFIBM-Growth (parallel frequent item binary matrix growth) algorithm. The algorithm constructs a binary matrix to compress and store data sets, and it uses simple matrix operations to mine frequent itemsets in parallel. In addition, we have applied this algorithm to the failure prognostic of primary air fans. We find that not only the algorithm is not bad, but also the association rules got from it could realize the failure prognostic of primary air fans.