Personalized Recommendation in Online Health Communities with Heterogeneous Network Mining

There is an increasing attention of Online Health Communities (OHCs) where health consumers exchange informational and emotional support from their peers. However, the information overload issue makes it difficult for e-patients to identify relevant threads for their needs. An effective thread recommendation approach is highly desirable for OHCs to improve user experience. In this work, we propose to represent OHCs data as a Heterogeneous Healthcare Information Network (HHIN). We extract node-based and path-based features, and train a binary classification model for personalized thread recommendation. We conduct an experiment using a dataset from a popular OHC, and the results show that our approach is promising in predicting users' preference in threads.