A deep learning approach integrating multi-dimensional features for expert matching in healthcare question answering communities

To address the demand for precise patient-medical expert matching in online healthcare Q&A communities, this study proposes a multi-feature health community expert recommendation model integrating GRU, convolutional neural networks (CNN), and attention mechanisms. By analyzing textual semantic features from patients' question titles, content, tags and personal profiles, while incorporating medical experts' professional credentials information and historical reply sequences, we construct a recommendation framework with multi-dimensional feature fusion. The CNN model extracts deep semantic information from patient inquiries, coupled with a bidirectional GRU network to align with experts' specialized medical domains, thereby optimizing recommendation accuracy and relevance. Experimental results demonstrate significant improvements in recommendation precision compared to traditional text matching methods (e.g., LSTM) and previous state-of-the-art approaches, particularly in handling unstructured, short-text, and multi-domain classification scenarios. This research provides technical references for resource optimization and personalized services in online medical communities, offering practical implementation value.