Enhancing health information recommendation with social network features: Overlapping communities and user influence

Health recommender systems (HRSs) are vital for helping users search for tailored information in online healthcare communities (OHCs). Social networks can reflect user latent features and improve recommendation accuracy; however, existing HRSs often overlook their potential. To address this issue, we propose a novel social health recommender system (SHRS) that incorporates overlapping communities and social influence in social networks. Using the community overlap propagation algorithm (COPRA) and LeaderRank algorithm, we analyse overlapping communities and user influence to better uncover latent features of users. A novel social regulation mechanism is then proposed, combining social network features with the rating matrix, and integrated into a matrix factorisation framework for recommendation. Experiments on a diabetes OHC data set show our model outperforms baseline models, achieving improvements of 12.8% in precision, 12.2% in recall and 12.9% in F 1-measure. This study advances HRSs by leveraging social network features, offering practical insights for researchers and OHC platforms.