An algorithm for mining frequent closed itemsets with density from data streams

Abstract

Mining frequent closed itemsets from data streams is an important topic. In this paper, we propose an algorithm for mining frequent closed itemsets from data streams based on a time fading module. By dynamically constructing a pattern tree, the algorithm calculates densities of the itemsets in the pattern tree using a fading factor. The algorithm deletes real infrequent itemsets from the pattern tree so as to reduce the memory cost. A density threshold function is designed in order to identify the real infrequent itemsets which should be deleted. Using such density threshold function, deleting the infrequent itemsets will not affect the result of frequent itemset detecting. The algorithm modifies the pattern tree and detects the frequent closed itemsets in a fixed time interval so as to reduce the computation time. We also analyse the error caused by deleting the infrequent itemsets. The experimental results indicate that our algorithm can get higher accuracy results, and needs less memory and computation time than other algorithm.