Graph Mining BGU图挖掘的BGU.pptVIP

* one node for each unique term: unique labeling of nodes – no two nodes have the same label * Second, smart method assume that relevant for classification feature is that frequent in one category and unfrequent in others. This is better assumption then we used in na?ve case. We developed measure called Classification Rate that give higher weight to such subgraphs. Extraction process is pretty similar to na?ve process. CR min is used as threshold. * * Vector models are represented by dotted lines, graph models are represented by solid lines Additional Approaches for Single Graph Setting BFS Ap