In the analysis of social network, the attribute values of an entity change constantly as time goes by and the corresponding attribute graph changes accordingly. However, the essence of the entity is invariable. The problem is how to discover essence from the change of mining frequent rough matching attribute sub-graph in an attribute graph in this paper. The problem of the attribute sub-graph rough matching is defined and described, then the random walk rough matching algorithm of attribute sub-graph is designed so that the problem of incomplete coincide attribute sub-graph rough matching would be solved. Attribute sub-graph is the expansion of the traditional sub-graph; rough matching problem is the extension of traditional sub-graph matching problem. With the help of the random walk rough matching algorithm of attribute sub-graph, more potential frequent attribute sub-graphs can be discovered, and more valuable information mined.