基于卷积神经网络的中草药推荐系统  

Convolutional Neural Network-Based Recommendation for Chinese Herbal Prescriptions

作　　者：陈灿宇 付阿敏 李国伟 余兆钗 李佐勇[1,3] Chen Canyu;Fu Amin;Li Guowei;Yu Zhaochai;Li Zuoyong(College of Computer and Control Engineering,Minjiang University,Fuzhou 350121,China;College of Electronic and Information Engineering,Shandong University of Science and Technology,Qingdao 266590,China;Fujian Provincial Key Laboratory of Information Processing and Intelligent Control,Minjiang University,Fuzhou 350121,China)

机构地区：[1]闽江学院计算机与控制工程学院,福州350121 [2]山东科技大学电子信息工程学院,青岛266000 [3]福建省信息处理与智能控制重点实验室(闽江学院),福州350121

出　　处：《世界科学技术-中医药现代化》2022年第6期2402-2410,共9页Modernization of Traditional Chinese Medicine and Materia Medica-World Science and Technology

基　　金：国家自然科学基金委员会面上项目(61972187):基于深度学习的血液白细胞分割与分类研究,负责人:李佐勇;福建省自然科学基金委员会重点项目(2020J02024):基于深度神经网络的血液白细胞分割与分类研究,负责人:李佐勇;福州市科技局科技计划项目(2020-RC-186):血液白细胞图像自动分析关键技术研究及软件系统开发,负责人:李佐勇。

摘　　要：中医诊疗在我国历史悠久,但是从计算机角度对中医症状与中药关系的研究很少。本研究基于中医诊疗规则,运用卷积神经网络技术,实现从中医病症到中草药的自动推荐(CNN-based Herb Prescription,CNN-HP)。具体地,本研究提出一种含单个卷积层和三个全连接层的卷积神经网络模型,并与现有中草药推荐算法以及经典的机器学习算法进行了推荐性能的比较。定量和定性的实验结果表明,CNN-HP模型的推荐性能高于对比算法,其精确度为71.54%,召回率为87.09%,F1分数为78.55%。本研究为中医病症到中草药的自动推荐提供了新思路。Traditional Chinese Medicine(TCM)diagnosis and treatment have a long history in China.However,there is little researches on the relationship between TCM symptoms and Chinese herbs from the perspective of computer.Based on the thought of TCM,this study used convolutional neural network to realize automatic recommendation from TCM symptoms to Chinese herbs(CNN-HP).Specifically,this study proposed a convolutional neural network model with a single convolutional layer and three full-connection layers,and compared its recommendation performance with the existing Chinese herbal recommendation algorithms and several classical machine learning algorithms.The quantitative and qualitative experimental results showed that the recommendation performance of our CNN-HP model was higher than those of the comparison algorithms,with an accuracy of 71.54%,a recall rate of 87.09%,and an F1 score of 78.55%.This study provided a new idea for the automatic recommendation of TCM symptoms to Chinese herbs.

关 键 词：中医辅助诊疗 人工神经网络 处方推荐 

分 类 号：R2-03[医药卫生—中医学]