AFIF: Automatically Finding Important Features in Community Evolution Prediction for Dynamic Social Networks
We present AFIF, an efficient solution to examine communities' structural features and also to find a proper subset of promising features in order to predict the upcoming changes of social networks.
In this work, we study social networks at a community structure level to predict communities' evolution over time.
AFIF chooses an appropriate subset of features in prediction of community evolution while maintaining an impressive performance.
Community evolution prediction: (a) dividing the social network into time windows, (b) detecting communities in each time window, (c) utilizing a community tracker for community matching and identifying chains of community evolution, and (d) concepts of features, output variable, and time window t (the youngest time window) in training a classifier.