An Optimistic Messaging Distributed Algorithm for Association Rule Mining

We present a distributed algorithm based on dynamic itemset counting (DlC)for generation of frequent itemsets. The nature of DIC represents a paradigm shift from Apiori-based algorithms in the number of passes of the database hence reducing the total time taken to obtain the frequent itemsets. Our distributed algorithm, called the optimistic messaging DIC (OPT-DIC) gives much better results than count distribution (CD). We exploit the advantage of dynamic itemset counting that of starting the counting of an itemset as early as possible. Hence, our algorithm shows remarkable improvement in the amount of time taken because of reduction in the number of passes of the database and comparatively lesser number of candidates generated.