On feature prediction in temporal social networks based on artificial neural network learning

The study of network features is an important analysis method for the social networks, and prediction of network features is a research problem with many applications, particularly in decision making. In this paper, we propose a novel feature prediction method for temporal social networks, which estimates network measurements in the future based on a small window of measurements in the past. We utilized artificial neural networks as a supervised learning algorithm for training the estimation functions. The comprehensive evaluations show that the proposed method outperforms alternative baselines remarkably according to the prediction accuracy.