Sulyun Lee, Hankyu Jang, Kang Zhao, Michael S. Amato, and Amanda L. Graham
Abstract: Social networks often incorporate multiple types of social relationships, making them multi-relational networks. Effective link predictions can help social networks improve user experience and engagement, but limited attention has been paid to predicting links in multi-relational networks. This paper explores link predictions in multi-relational networks from an online health community. We demonstrate that leveraging information from multiple networks built based on different types of relationships is superior to using only information from a single network or the aggregated network. In addition, adding community structures, nodal similarities based on network embedding and topic similarity can help link predictions in different ways. Our work has implications for the design and management of a successful online health community.
Paper link: PDF
Sulyun Lee, Hankyu Jang, Kang Zhao, Michael S. Amato, and Amanda L. Graham. Multi-Relational Link Prediction for an Online Health Community. INFORMS Data Science Workshop, 2019.