Feature Selection: Evaluation, Application, and Small Sample Performance
IEEE Transactions on Pattern Analysis and Machine Intelligence
gSpan: Graph-Based Substructure Pattern Mining
ICDM '02 Proceedings of the 2002 IEEE International Conference on Data Mining
Direct mining of discriminative and essential frequent patterns via model-based search tree
Proceedings of the 14th ACM SIGKDD international conference on Knowledge discovery and data mining
Graph classification based on pattern co-occurrence
Proceedings of the 18th ACM conference on Information and knowledge management
Forward semi-supervised feature selection
PAKDD'08 Proceedings of the 12th Pacific-Asia conference on Advances in knowledge discovery and data mining
Hi-index | 0.00 |
In the real world, there exist many graph applications that have only a small number of labeled graph data. Graph classification can achieve desired performances only if enough amounts of labeled data are available. Semi-supervised feature selection studies have been proposed to solve the short labeled data problem. However, the existing studies still do not meet satisfactory classification accuracy. In this paper, we propose a novel semi-supervised feature selection method, named co-occurrent graph feature selection (SCGFS) that selects a set of optimal co-occurrent frequent subgraph features for graph classification. Co-occurrent subgraphs can have higher discriminative powers than those of individual frequent subgraphs. Since co-occurrent patterns have an inefficiency problem, we propose a branch-and-bound search that reduces the feature search space with effective pruning techniques. Through comprehensive experiments, we show that the proposed framework can efficiently select more discriminative subgraph features compared with existing semi-supervised feature selection algorithm.