An Approximate Non-Derivable Itemset Mining Algorithm over Data Streams

Abstract

Frequent itemset mining is one of the traditional and important problems in data mining.Non-derivable frequent itemsets are the condensed reprentation of frequent itemsets,and they can not only reduce the memory cost,but also make association rules more understandable for user.Because the bound computations of non-derivable frequent itemsets are high,the authors propose the conception of approximate non-derivable itemsets,and present an approximate non-derivable frequent itemset mining algorithm MANDI based on itemset idlist.In addition,the authors present the stream mining algorithm UNADI,which maintains the negative borders of approximate non-derivable frequent itemsets to conduct efficient incremental mining.The experimental results show that both algorithms are effective and efficient.