Discriminating Frequent Pattern Based Supervised Graph Embedding for Classification

Graph is used to represent various complex relationships among objects and data entities. One of the emerging and important problems is graph classification that has tremendous impacts on various real-life applications. A good number of approaches have been proposed for graph classification using various techniques where graph embedding is one of them. Here we propose an approach for classifying graphs by mining discriminating frequent patterns from graphs to learn vector representation of the graphs. The proposed supervised embedding technique produces high-quality entire graph embedding for classification utilizing the knowledge from the labeled examples available. The experimental analyses, conducted on various real-life benchmark datasets, found that the proposed approach is significantly better in terms of accuracy in comparison to the state-of-the-art techniques.