Link Weight Prediction with Deep Learning
MetadataShow full item record
Deep learning has been successful in various domains including image recognition, speech recognition and natural language processing. However, the research on its application in graph mining is still in an early stage. Here we present Model R, a neural network model created to provide a deep learning approach to the link weight prediction problem. This model uses a node embedding technique that extracts node embeddings ( knowledge of nodes) from the known links' weights (relations between nodes) and uses this knowledge to predict the unknown links' weights. We demonstrate the power of Model R through experiments and compare it with the stochastic block model and its derivatives. Model R shows that deep learning can be successfully applied to link weight prediction and it outperforms stochastic block model and its derivatives by up to 73\% in terms of prediction performance. We analyze the node embeddings to confirm that closeness in embedding space correlates with stronger relationships as measured by the link weight. We anticipate this new approach will provide effective solutions to more graph mining tasks. We also demonstrate how to apply Model R for the collaborative rating prediction problem, which is a special case of the graph link weight prediction problem where the graph is a bipartite graph. This model extracts knowledge of users and items from known ratings and uses this knowledge to predict unknown ratings. We demonstrate the power of Model R through experiments and compare it with the most prevalent collaborative filtering algorithm - neighborhood-based collaborative filtering. Model R shows that deep learning can be successfully applied to collaborative rating prediction and it outperforms neighborhood-based collaborative filtering by up to 18\% in terms of prediction performance. We also present Model S, a generalized deep learning approach to the graph link weight prediction problem based on general node embedding techniques. We evaluate this approach with three different node embedding techniques experimentally and compare its performance with stochastic block model and its derivatives.