Cooperation Prediction In Github Developers Network With Restricted Boltzmann Machine

In order to solve link prediction problem with higher accuracy than achieved by classical supervised approaches we provide a proposal of the method based on information extracted from network using pre-processing done by Restricted Boltzmann Machine (RBM) and statistical inference models. Input space is fed to RBM in order to provide new sparse coded feature space that is used in order to estimate parameters of classical inference models. By accomplishing link prediction with proposed RBM pre-processing noticeable increase of all accuracy related measures was observed in comparison to state-of-the-art approaches.