Coupled Semi-supervised Clustering: Exploring Attribute Correlations in Heterogeneous Information Networks.

Heterogeneous Information Network (HIN) has been widely adopted in various tasks due to its excellence in modeling complex network data. To handle the additional attributes of nodes in HIN, the Attributed Heterogeneous Information Network (AHIN) was brought forward. Recently, clustering on HIN becomes a hot topic, since it is useful in many applications. Although existing semi-supervised clustering methods in HIN have achieved performance improvements to some extent, these models seldom consider the correlations among attributes which typically exist in real applications. To tackle this issue, we propose a novel model SCAN for semi-supervised clustering in AHIN. Our model captures the coupling relations between mixed types of node attributes and therefore obtains better attribute similarity. Moreover, we propose a flexible constraint method to leverage supervised information and network information for flexible adaption of different datasets and clustering objectives. Extensive experiments have shown that our model outperforms state-of-the-art algorithms.