一种基于滑动窗口的多关系模式频度更新算法

面向多个相关数据流的挖掘算法研究尚处于起步阶段。作为多数据流挖掘算法的基础,模式频度更新算法仍然存在计数不准确、性能较低等问题,难以以此构造有效的挖掘算法。通过引入多关系挖掘概念以及目标关系定义,进而限定计数对象,提出了一种基于滑动窗口的多关系模式频度更新算法MRPFU。该算法监视各数据流窗口的更新情况,采用计数传播策略,减少了时间与空间复杂度。理论分析及实验结果证明了所提算法的有效性且具有较高性能。

Multi-relational pattern frequency update algorithm based on sliding window

Presently,the study of mining algorithms for multiple correlated data streams is still at a primitive stage.As the basis of mining multiple data streams,the methods of updating the frequencies of patterns,are bearing problems of count deviation,low performances etc.Consequently,efficient mining algorithms are difficult to be built either.The concepts of multi-relational data mining and target relation are introduced firstly,and the count object is defined accordingly.Then an algorithm based on sliding windows for updating frequencies of multi-relational patterns is proposed,which monitors the updates of streams,adopts the strategy of count propagation,and relieves the complexity of runtime and space.The theoretical analysis and experiments prove its effectiveness and performance.