| In recent years, the problem of frequent itemset mining has evolved into a new form called data stream mining. In it, a continuous stream of data is mined or analyzed to find frequent itemsets. The continuous nature of streams brings new challenges to (traditional) frequent itemset mining. Although some algorithms have been proposed to mine frequent itemsets from data streams, they do not provide user control over the mining process through the use of constraints. Constraint-based mining can achieve the goal of user exploration. Moreover, various pruning techniques for constraint-based algorithms can be utilized in the stream mining process to make it more efficient.; In this thesis, we develop an algorithm for constraint-based frequent itemset mining from data streams. Specifically, we propose a novel tree structure, called Data Stream Storage Tree (DSSTree), to efficiently store and process incremental updates of streaming data. Experimental results show that our algorithm enables users to have a better control on the mining process and to extract interesting frequent itemsets from continuous streams of data efficiently. |
