Skip to Main Content
Most recommendations are made based on the computation of user specified constraints or functions of object similarity. In this paper, we discuss a new trend of recommender systems that are based on the information network analysis to exploit the relationships between data objects. An information network can be constructed from different application networks such as social media, traffic management systems, and sensor networks. Heterogeneous information networks are now ubiquitous. How to make recommendations for the evidence-based decisions based on the fusion of these information networks presents a challenge. We present a framework of recommendations based on information network analysis. The practical examples are used to demonstrate the potential of this type of recommendation techniques. The evaluation methodologies for network-base recommendations are also addressed.