Abstract: Aiming at the problem that none of current researches on the WMFI (weighted maximal frequent itemsets) over data stream emphasizes the WMFI mining on the condition that the frequent threshold is not equal with the weighted frequent threshold， the concept of FWMFI (full weighted maximal frequent itemsets) was firstly promoted in this work. In order to reduce redundant operations existing in the naive algorithm which is used to handle the FWMFI mining based on sliding window over data stream， the FWMFI-SW (FWMFI mining based on sliding window over data stream) algorithm was proposed. The mining optimization strategy was adopted based on the frequent character to reduce the unnecessary call about the MaxW optimization strategy in the naive algorithm. In addition， the edit distance ratio was taken as reconstruction judge function to decide whether the updated WMFP-SW-tree should be reconstructed as the window slides. The extensive experiments showed that the FWMFI-SW algorithm is effective ， and outperforms the naive algorithm in running time.

Key words: data stream, sliding window, edit distance ratio, weighted maximal frequent itemsets, reconstruction judge function