GAT-Based Rating Model for Fair Recommendations on Online Health Communities

Online Health Community (OHC) is a platform that provides medical consultation and health knowledge-sharing services. For patients, OHC can recommend high-quality doctors according to their recommendation popularity (RP). Current OHC relies on factors such as visit frequency and consultation volume to rate RP primarily. However, this rating method will make doctors with high RP more likely to be discovered and selected by patients, further increasing their RP. This amplifies the popularity bias of doctors in each rating. In this paper, a rating model is proposed based on graph attention networks (GAT) to address this problem. This model makes full use of the characteristics and relationships of doctors to create similar trends in the RP of similar doctors by using graph data. Through data analysis, this method reduces the differences between high-RP and low-RP doctors, leading to a more balanced distribution of RP on OHCs and effectively limiting popularity bias.