Abstract:

Based on the algorithm of DSMMFI, an improved algorithm, named DSMMFIDS (Dictionary Sequence Mining Maximal Frequent Itemsets over Data Streams), is proposed. Firstly, it stores transaction data into DSFIlist in alphabetical order. Secondly, the data are stored sequentially into the tree similar to the summary data structure. Thirdly, nonfrequent items in the tree and DSFIlist are removed, and the transaction items with the maximum count of window attenuation supports are deleted. Finally, the strategy (topdown and bottomup twoway search) is used to mine maximal frequent itemsets over data streams, and case analysis and experiments prove that the algorithm DSMMFIDS has better performance than the algorithm DSMMFI.
    

Key words: data mining;data stream;landmark windows;maximal frequent itemsets;window attenuation support count