关键词: 频繁子图，图分类，图挖掘，特征选择

Abstract: Many graph classification approaches have been proposed. These approaches only look at the structural information of the pattern, but do not take advantage of the embedding information during mining frequent subgraph. In facts,in some efficient subgraph mining algorithms,the embedding information of a pattern can be maintained. A graph classification approach was presented. Based on L-CLAM coding,it uses a feature subgraph selection strategy based on label information to select the feature subgraph, while making full use of embedding sets to directly generate feature subgraph in mining frectuent subgraph. Experiment results show that it is effective and feasible.

Key words: Frequent subgraph, Graph classification, Graph mining, Feature selection