一种融合标签和患者咨询文本的医生推荐算法

A Physician Recommendation Algorithm Based on the Fusion of Label and Patient Consultation Text

【目的/意义】针对在线医疗信息结构松散,医疗平台医生推荐精度不足的现状,设计了一种基于标签和患者咨询文本的医生推荐算法,提升医生推荐效果。【方法/过程】利用Word2vec模型训练患者咨询文本得到特征向量,改进余弦相似度算法计算医生推荐集A;利用LDA模型训练医生标签得到医生在主题上投影的概率分布,改进KL距离算法计算医生推荐集B;基于社会网络分析理论设计相关算法重构医生网络链接,选择中心性指标得到最终医生推荐集C。【结果/结论】以“丁香医生”数据进行实证,面向UGC数据丰富了算法的可用程度,弥补了单一推荐方法的不足,提高了推荐的精度。本文所提方法有效提升了医生推荐精度。【创新/局限】通过融合标签和患者咨询文本,采用社会网络分析实现了医生混合推荐。虽然通过中心性指标进行重要医生挖掘,但挖掘效果有提升空间。

【Purpose/significance】Aiming at the current situation of loose structure of online medical information and insufficient precision of online physician recommendation,a physician recommendation algorithm based on label and patient consultation text is designed to improve the effect.【Method/process】Word2vec model is used to train patient consultation texts to obtain feature vectors and improve cosine similarity algorithm to calculate physician recommendation set A;LDA model is used to train labels to obtain probability distribution of projection on the subject and KL distance algorithm is improved to calculate set B;based on social network analysis theory,relevant algorithms are designed to reconstruct physician network links and select centrality index to obtain final set C.【Result/conclusion】Based on the relevant data of "doctor clove",the non-traditional UGC data enriches the availability of the algorithm,makes up for the deficiency of single recommendation method,and improves the accuracy.The proposed method effectively improves the accuracy of doctor recommendation.【Innovation/limitation】By fusing label and patient consultation text,social network analysis is used to realize physician hybrid recommendation.Though the central indicators of important doctors mining,the effect has room to improve.