Link Prediction-Based Multi-label Classification on Networked Data

In this paper, we study the problem of performing multi-label classification on networked data, where each instance in the network is assigned with multiple labels and the connections between instances are driven by various casual reasons. Networked data extracted from social media or web pages may not reflect the relationship between users in real life accurately. By mining the links that actually exist but have not yet been found in the network, the potential relations between users can be discovered, and thus help us to predict the users' labels more accurately. In this work, we propose a link prediction-based multi-label relational neighbor classifier which employs social context features (LP-SCRN). It firstly predicts missing links in the network, and then calculates the weights of the links according to the similarity between nodes in their social features. In addition, by capturing the potential correlation between nodes, we expand a node's neighbor set, and refine the multi-label relational classifier. Experiments on two real-world datasets demonstrate that our proposed method improves the performance of multi-label classification on networked data.