Personalized Doctor Recommendation Based On Text Analysis

Through the analysis of the dataset obtained by the network crawler,we can fully mine the information in it.We can analyze the doctor-patient interaction behavior and extract the individual characteristics of the doctors and patients and then recommend the doctors.In this way,we can provide efficient,accurate and convenient doctor service when the patients carry on the online inquiry.The doctor’s recommendation of the text mainly includes two parts.The first part is mainly based on the similar patients,and the second part is mainly based on the project-based collaborative filtering algorithm.Firstly,the paper uses the text analysis technology——word2vec model to vectorize the patient consultation text and the doctor diagnosis suggestion text,so as to build their characteristic models.In the first recommendation algorithm,we cluster patients firstly.Then we recommend separately in each category.We form a preliminary doctor candidate based on the similarity between patients.Then we sort the doctor candidate set by combining the doctor features,where we choose the doctor feature information based on the text of the patient consultation.In another recommendation algorithm,the project-based collaborative filtering algorithm is mainly used.In order to make the the algorithm get a good effect,we solve the problem of how to choose multiple doctors by merging a few similar patients.In addition,the patient-to-user scoring dataset is obtained by the doctor’s characteristic information based on the diagnostic advice and the doctor’s scoring indexes,which is a very important dataset in collaborative filtering algorithm.In the end,the experiments’ results of the two recommendation methods have high accuracy.And compared with the actual situation,the doctor’s recommendation has also achieved very good results.In addition,the two recommendation algorithms think from different angles,the information used is not exactly the same,and the processing methods are also very different.So we choose to fuse their results for mixed recommendation.Finally through the mixed model,our doctors-recommend can achieve higher accuracy and better result.