Published Paper Details:

Mining a frequent itemsets from databases is really hard task to perform. This can be done using a traditional parallel algorithms .Using these algorithms itemsets are equally partitioned among a group of computing nodes. But these algorithms gives a performance issues like high communication and mining overhead induced by redundant transactions transmitted among computing nodes, database has to scan multiple times and huge candidate keys needs to be exchanged between the processor, I/O and synchronization, impracticableness to construct in-memory FP trees to manage large databases. To solve all these problems we have implemented FiDoop a parallel frequent itemset mining algorithm. FiDoop develops frequent items ultrametric tree instead of conventional FP trees to increase the performance of algorithm. It uses three MapReduce jobs and develops a workload balance metric to measure load balance across the nodes of cluster. In this way FiDoop helps to improve the parallel mining of frequent data itemsets by incorporating frequent items ultrametric trees.