基于普通ER模型的多关系序列模式挖掘算法

Multi-relational Sequence Patten Mining Based on Normal ER Model

在普通ER模型下,为了避免多表物理连接,本文提出了一种挖掘多关系序列模式的算法(MSPEM).该算法首先指定目标表及关键原子,建立相应的语义关系图,再根据语义关系图中的连接路径,传播元组标号.挖掘过程可分为单表挖掘和跨表挖掘,支持度的计算以关键原子为依据进行统计,最终得到基于虚拟连接表的多关系序列模式.MSPEM是一种ER模型下的基于元组传播的多关系序列模式挖掘方法,同时通过指定关键原子解决了基于虚拟连接表存在的统计偏斜问题.实验结果表明MSPEM算法具有更高的执行效率.

In order to avoid physical link in normal ER model , this paper proposes a method for mining multi-relational sequence pattern（ MSPEM ）. In this Algorithm, the target table and the key element are firstly assigned, and the semantic relationship graph is established. , then tuple labels are propagated along with the connected paths in the graph. Mining process can be divided into single-table mining and cross-table mining, and the support is calculated based on key element, eventually multi-relational sequence patterns are obtained based on the virtual connection table. MSPEM is a way to mine multi-relational sequence sequence pattern in ER model based on tuple propagation, which can solve the statistics decline by assigning key element based on virtual table connection. Experimental result shows that MSPEM has higher efficiency.