【摘要】 在许多场合挖掘频繁闭合序列时,输入串数据库呈现实时动态增长的特点.分析Bide算法,给出并证明了闭合序列前缀中任意一个项目的后向扩展事件(BEE)项目交集随前缀的生长单调不增的定理,据此对BEE累计操作进行了优化,使其性能平均提高了4.8%.定义了闭合序列树作为频繁闭合序列的表示形式,并阐述了它的3个性质.分析发现,当新增输入串不同时包含前缀串和频繁项目时,两次连续挖掘的结果是相同的,给出了相应的定理和证明,据此实现了增量式频繁闭合序列挖掘算法BideInc.实验验证了BideInc算法的正确性,使用该算法后挖掘性能平均提高了47%.更多还原

【Abstract】 While mining frequent closed patterns(FCP),the input sequence database dynamically increases in many situations.By analyzing Bide algorithm,the theorem of backward-extension event(BEE) detection was proposed and proved.It shows that the BEE set of any prefix item is non-increasing with the extension of the prefix.Based on the theorem,the accumulation performance of the BEE set was optimized by 4.8% averagely.The FCP tree was defined to represent the final result of FCP mining and its three characteristics were demonstrated.When the frequent item and the prefix are not coexistent in the new input sequence,the results of contiguous FCP mining are equal.And the corresponding theorem was proved.The BideInc algorithm was proposed to incrementally mine FCPs.The experiments validated the algorithm,and the performance was improved by 47% averagely.更多还原