A Study on Doctor Recommendation Model in Medical Guidance Services

In medical guidance services, it is of great significance to match the appropriate department and doctor by digging deeper into patients’ demands. However, accurate matching of doctors requires the ability to locate the exact department based on the text of the patients’ chief complaint, and then select the matching doctor by considering the patients’ condition, the doctor’s professionalism, and the patients’ preference. To this end, this paper proposes a department classification model on the basis of Convolutional Neural Networks (CNN) as well as Robustly optimized BERT approach (RoBERTa) with an attention mechanism. The model firstly extracts the patients’ chief complaint texts features by convolution layer, and then introduces the attention mechanism to assign different weights to different features. Subsequently, these features are fused with the features extracted by RoBERTa for classification. In addition, this paper proposes a doctor recommendation algorithm that considers both patient similarity and patient preference. Through the in-depth analysis on the patients’ condition claims, various weights are assigned to various influencing factors, and then the matching degree is calculated to achieve the accurate recommendation of doctors. The experimental results reveal that the proposed department classification model’s accuracy on the dataset is 93.4%, and the Normalized Discounted Cumulative Gain (NDCG) of the doctor recommended algorithm is 90.7%. In this way, the proposed model effectively improves patient-doctoral matching with excellent performance.