Algorithm Data Stream Frequent Pattern Mining

The data stream is a high speed and unlimited arrival ordered itemsets. It is different from thedata in the traditional static database. The characteristics of the data stream is continuous,real-time, unlimited and so on. And these data arrival with a fast rate.Since the data stream issignificantly different from the static data.Therefore, these algorithms of mining frequent itemsetson data mining has existed are difficult to apply in the data streams. Now there are a lot ofscholars research in the algorithm of mining frequent itemsets over data streams.Mining frequentitemsets on data streams has become one of the main problems in data mining.This paper describes the state of mining data streams at domestic and foreign. This paperdiscusses the application of data mining and data mining technology, the main problem of miningfrequent pattern on data streams and the window mechanism of data streams.Algorithms of DMF-MFI is mining frequent itemsets about historical transaction over datastreams.Algorithms of MS is mining frequent pattern within any sliding time window.Algorithmsof MFI-TransactionSW is mining transaction data’s frequent itemsets in the sliding window.Theseare existing algorithm of mining data streams’ frequent pattern. This article analyzes andsummarizes these algorithms.The article proposed the algorithm of MFI-TransactionLW of mining frequent itemsets inlandmark window based on the the MFI-TransactionSW algorithm. This algorithm uses the thebitmap structure stored items, constantly updated list of BSIR-list and create frequent pattern treeBSFP-tree. Through top-down search strategy for mining frequent itemsets. This article comparesthe the algorithm of MFI-TransactionSW and algorithms of MFI-TransactionLW through use caseanalysis. Finally, the algorithm of MFI-TransactionLW takes less time than the algorithm ofMFI-TransactionSW and require less memory space by experiments.Finally, this paper proposed DSMMFI-DS algorithm, the algorithm sort transaction inDSFI-list according to a certain order of the total order, and then stored sequentially in sorted to asimilar summary of the data structure tree of DSSEFI-tree.Second,pruning non-frequent itemsand the itemsets of the number of window attenuation support very small from DSFI-list. Finally,this paper make use of the strategy(top-down and bottom-up two-way search) to mining maximalfrequent itemsets over data streams.