Multi-label Collective Classification Using Link Based Label Diffusion

Abstract

Due to the growth of numerous applications of network mining, node classification has become an important task on various domains like social networks, biological networks and communication networks. Exploiting the dependencies or correlations among the nodes in the network is a major challenge in node classification. Performing classification based on these correlations is known as collective classification. Classification problem where each observation can have multiple target labels is referred to as multi-label classification. The correlation between these multiple target labels makes multi-label classification a difficult task. In this paper, we address the problem of multi-label collective classification, which has to consider both between-node and between-label correlations. In this work, we propose a novel method for multi-label collective classification using link based label diffusion (MCLD), which exploits both the structural properties of network and label correlations among the nodes. We conducted experiments using various network datasets. We evaluated the efficiency of the system using various evaluation measures and compared the results with the state-of-the-art methods.