Approximately Mining Recently Representative Patterns on Data Streams

Catching the recent trend of data is an important issue when mining frequent itemsets from data streams. To prevent from storing the whole transaction data within the sliding window, the frequency changing point (FCP) method was proposed for monitoring the recent occurrences of itemsets in a data stream under the assumption that exact one transaction arrives at each time point. In this paper, the FCP method is extended for maintaining recent patterns in a data stream where a block of various numbers of transactions (including zero or more transactions) is inputted within each time unit. Moreover, to avoid generating redundant information in the mining results, the recently representative patterns are discovered from the maintained structure approximately. The experimental results show that our approach reduces the run-time memory usage significantly. Moreover, the proposed GFCP algorithm achieves high accuracy of mining results and guarantees no false dismissal occurring.