A Parallel AprioriAll-Based Sequential Pattern Mining Algorithm on MapReduce

AprioriAll is a well-known algorithm for sequential pattern mining, an important and very useful for identifying patterns that can be used for predicting behaviors and future trends to answer business questions. However, AprioriAll algorithm has some limitations concerning sequential pattern mining of huge datasets where the technique suffers in performance and scalability. There are some attempts to handle this problem in distributed way, And we as well in this paper proposes a parallel sequential pattern mining algorithm called PAA (Parallel AprioriAll). PAA is based on a AprioriAll algorithm with modifications to make use of a Concept Lattice model in order to run it in a distributed fashion using Hadoop and MapReduce. We have developed a prototype implementing the PAA algorithm and made a comparative study with stand-alone AprioriAll algorithm. The empirical comparison results show that PAA algorithm is correct, produce the same result and outperform the AprioriAll algorithm.