No abstract

A distributed algorithm based on Dynamic Itemset Counting (DIC) for generation of frequent itemsets is presented by us. DIC represents a paradigm shift from Apriori-based algorithms in the number of passes of the database hence reducing the total time taken to obtain the frequent itemsets. We exploit the advantage of Dynamic Itemset Counting in our algorithm- that of starting the counting of an itemset as early as possible at the different site as soon as they become frequent at atleast one site. 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. Distributed frequent itemset counting and association rule generation have basically used algorithms based on Apriori or Sampling. This is the first algorithm which is based on DIC.

No abstract