Doctor-Patient Compatibility in Online Healthcare Based on Dynamic Fusion of Multimodal Uncertainty Information

With the reopening post-pandemic and the advancement of smart healthcare, online consultations have gradually become one of the primary means of health management due to their convenience and low cost. However, the uncertainty in different modalities between patients and doctors has further complicated the challenge of matching. How to achieve efficient patient-doctor matching has become a pressing issue. Therefore, this paper proposes a multimodal uncertainty dynamic fusion method. First, we design a method for measuring uncertainty based on the modality information of each doctor and patient. Then, considering the upper bound of the model's error, we introduce a regularization term between modality weights and loss to improve the loss function. Finally, we conduct model comparisons and uncertainty information fusion analysis using the data provided by the Haodaifu online medical platform. The results demonstrate the effectiveness of the proposed method.