Abstract: Graph can express richer semantic meaning, but enhance the complexity of data structure and increase the difficulty of interested sub-graph mining. The knowledge of graph and data mining technology is applied comprehensively to accomplish canonical coding of graph. Candidate subgraph is generated by join and extension operation. The value of support and frequency of candidate subgraph is counted by maintaining an embedded set. Algorithm FSubgraphM is proposed to mine frequent induced subgraph from graph database.

Key words: data mining, subgraph isomorphism, canonical coding, embedded set, frequent subgraph mining

摘要： 图像能表达丰富语义，但增加了数据结构的复杂性和感兴趣子结构的挖掘难度。综合应用图论知识和数据挖掘的各种技术，对图像进行规范化编码，通过连接和扩展操作产生所有候选子图，引用嵌入集概念，计算候选子图的支持度和频繁度。提出频繁子图挖掘算法FSubgraphM，能从图数据库中挖掘频繁导出子图。

关键词: 数据挖掘, 子图同构, 规范化编码, 嵌入集, 频繁子图挖掘

CLC Number: 

• TP311.2