Feature selection algorithm based on the Community discovery

In order to overcome the SVM for text classification ignoring the context of semantic information and the use of a community to text classification, one boundary point can only belong to a community of view, the concept of contribution and overlapping coefficient based on the complex network diagram is introduced. And feature selection algorithm based on community discovery is proposed. Experiments show that the algorithm can remain the text in the context of semantic information, which is more reasonable to divide the boundary points into the community. Then in the Bayesian classifier validate the algorithm performance in the recall, precision and F1 values. The result shows that the algorithm performance is superior to MIDF, and has a good flexibility in text classification.