利用局部图关联信息挖掘加权频繁模式

Mining weighted frequent patterns using local graph linking information

遍历模式数据挖掘方法已经在多种应用中被提出,传统的遍历模式挖掘仅仅考虑了非加权遍历。为解决加权遍历模式挖掘问题,首先提出了一种从EWDG(边加权有向图)到VWDG(顶点加权有向图)的变换模型;基于这种模型,提出了在具有层次特性的局部图遍历中,挖掘加权频繁模式的LGTWFPMiner(局部图遍历加权频繁模式挖掘法)及其支持度/权值界的局部评估方法。针对合成数据的实验结果表明该算法能够有效地进行基于图遍历的加权频繁模式挖掘。

Data mining for traversal patterns has been found useful in several applications. However, traditional model of traversal patterns mining only considered unweighted traversals. This paper proposed a transformable model of EWDG （ edgeweighted directed graph） and VWDG （vertex-weighted directed graph）to resolve the problem of weighted traversal patterns mining. Based on the model ,developed a new algorithm ,called LGTWFPMiner（ local graph traversals-based weighted frequent patterns miner） , and its local estimation of support/weight-bound to discover weighted frequent patterns from the traversals on graph with a level property. Experimental results of synthetic data show the algorithm is effective to resolve the problem of mining weighted frequent patterns based on graph traversals.