Community Evolution Prediction Based on Multivariate Feature Sets and Potential Structural Features

Abstract

The current study on community evolution prediction ignores the problem of internal community topology characteristics and does not take feature sets extraction into account. Therefore, the MF-PSF (Multivariate Feature sets and Potential Structural Features) method based on multivariate feature sets and potential structural features for community evolution prediction is proposed in this paper. Firstly, the multivariate feature sets are built from four aspects: community core node features, community structural features, community sequential features and community behavior features. Secondly, the community’s potential structural characteristics based on DeepWalk and spectral propagation theories are extracted, and the overall community’s internal structural characteristics and vertex distribution are analyzed. Finally, the community’s multivariate structural features and potential structural features are merged to predict community evolution events, and the importance of each feature in the process of evolutionary prediction is discussed. The experimental results show that compared with other community evolution prediction methods, the MF-PSF prediction method not only provides a foundation for analyzing the influence of various feature sets on predicted events, but it also effectively improves the accuracy of evolution prediction.