Finding Itemset-Sharing Patterns in a Large Itemset-Associated Graph

Both itemset mining and graph mining have been studied independently. Here, we introduce a novel data structure, which is an unweighted graph whose vertices contain itemsets. From the data, we enumerate combinations of connected subnetworks whose vertices share itemsets because the combinations imply side effects of drugs. To solve the problem, we introduce a novel algorithm ROBIN, which first enumerates connected graphs sharing itemsets, and then enumerates the combinations of the graphs. Our experimental result shows that ROBIN can solve the problem from the graph having more than one million edges, and enumerate biologically meaning subgraphs from real yeast dataset. Key words Graph, itemset, gene expression

• Type: Conference
• Year: 2010
• Where: PAKDD
• Authors: Mutsumi Fukuzaki, Mio Seki, Hisashi Kashima, Jun Sese
• Keywords: Data Mining, Graph Mining, PAKDD 2010, Vertices Contain Itemsets, Vertices Share Itemsets