Analyzing navigation history of a web site and that of a user reported in log files and web click stream helps in determining the components that shall be needed in the near future and fetch them in advance. Identifying the frequently visited websites from such data enhances the process. Processing and analyzing the vast amounts of click stream data and web logs that are generated at very high rates in a smart and cost-efficient way is a daunting challenge to the data mining community. Mining frequent itemsets plays an important role in analyzing such data streams. In spite of the existence of many such mining algorithms, more time and space efficient algorithms for mining frequent patterns are the need of the hour and are attracting wide attention in web click-stream analysis in recent years. In this paper an effective algorithm for mining frequent itemsets from a time-sensitive sliding window is proposed. It is a one-pass algorithm that uses a circular queue to implement the sliding window. The time-sensitive sliding window stores the web click stream data of various sessions of the web users. The proposed one-pass algorithm, FIM_CQTimeSWin has three phases: representation of time unit, maintenance of the sliding window and generation of frequent itemsets in the current sliding window.
Mining Frequent Itemsets from Data Streams with a Time-Sensitive Sliding Window
