滑动窗口下数据流完全加权最大频繁项集挖掘

Mining Full Weighted Maximal Frequent Itemsets Based on Sliding Window over Data Stream

针对当前关于数据流加权最大频繁项集WMFI(weighted maximal frequent itemsets)的研究无法有效地处理频繁阈值和加权频繁阈值不一致情况下WMFI的挖掘问题,提出了完全加权最大频繁项集FWM FI(full w eighted maximal frequent itemsets)的概念.为了减少naive算法在处理滑动窗口下完全加权最大频繁项集挖掘时存在的冗余运算,提出了FWMFI-SW(FWMFI mining based on sliding window over data stream)算法.所提出的算法通过基于频繁约束条件的优化策略减少了naive算法中M ax W优化策略的无效调用次数;采用编辑距离比率作为WMFP-SW-tree的重构判别函数,可以有效减少该树的重构次数.实验结果表明FWMFI-SW算法是有效的,且比naive算法更有时间优势.

Aiming at the problem that none of current researches on the WMFI （ weightedmaximal frequent itemsets） over data stream emphasizes the WMFI mining on the condition thatthe 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 reduceredundant operations existing in the naive algorithm which is used to handle the FWMFI miningbased on sliding window over data stream, the FWMFI - SW （ FWMFI mining based on slidingwindow over data stream） algorithm was proposed. The mining optimization strategy was adoptedbased on the frequent character to reduce the unnecessary call about the MaxW optimizationstrategy in the naive algorithm. In addition, the edit distance ratio was taken as reconstruction9udge function to decide whether the updated WMFP - SW - tree should be reconstructed as thewindow slides. The extensive experiments showed that the FWMFI - SW algorithm is effective,and outperforms the naive algorithm in running time.