中文摘要：

频繁子图挖掘是图挖掘的一个重要研究课题．gSpan算法作为一种高效的子图挖掘算法具有较好的执行效率，它通过最右扩展生成频繁子图，但不能保证每次扩展得到的均为标准编码．针对此问题本文提出了一种改进的算法CSGM，它采用ADI＋＋存储结构，能处理更大规模的图集，同时保证每次最右扩展均生成标准编码，既避免了对非标准编码图的支持度计算，也避免了对输入编码是否为标准编码的计算．在实际数据集上运行的实验结果表明它比原算法提高了挖掘效率．

英文摘要：

Frequent graph mining is an important research subject of graph mining. As an efficient algorithm for subgraph mining, the gSpan has better efficiency through fight-most extension to generate frequent subgraph. But it can＇t guarantee each extension is a canonical code. An improved algorithm called CSGM is proposed in this paper to resolve the existing problem. It can deal with larger graph dataset through using the ADI ＋ ＋ storage structure and guarantee that each extension is a canonical code. The advantage is that it not only avoids the calculation of the support of the non-canonical code graphs but also the calculation of whether an input code is a canonical code. The experimental result on real datasets shows that the new algorithm improves the efficiency of mining.