Recently, the data stream, which is an unbounded sequence of data elements generated at a rapid rate, provides a dynamic environment for collecting data sources. It is likely that the embedded knowledge in a data stream will change quickly as time goes by. Therefore, catching the recent trend of data is an important issue when mining frequent itemsets from data streams. Although the sliding window model proposed a good solution for this problem, the appearing information of the patterns within the sliding window has to be maintained completely in the traditional approach. In this paper, for estimating the approximate supports of patterns within the current sliding window, two data structures are proposed to maintain the average time stamps and frequency changing points of patterns, respectively. The experiment results show that our approach will reduce the run-time memory usage significantly. Moreover, the proposed FCP algorithm achieves high accuracy of mining results and guarantees no false dismissal occurring.