Incremental Frequent Itemsets Mining with MapReduce

Abstract

Frequent itemsets mining is a common task in data mining. Since sizes of today’s databases go far beyond capabilities of a single machine, recent studies show how to adopt classical algorithms for frequent itemsets mining for parallel frameworks such as MapReduce. Even then, in case of a slight database update a re-run of the MapReduce mining algorithm from the beginning on the whole data set is required and could be far from optimal. Thus, a variation of these algorithms for incremental database update is desired.