Traditional Chinese Medicine Recommendation Based On Deep Neural Network

Tongue diagnosis is a unique and important part of Chinese medicine.It is an important method to assist diagnosis by observing changes in the color and shape of the tongue.In this paper,through modern computer technology,the detection and analysis of tongue coating images,combined with traditional Chinese medicine theory and expert experience,provide accurate,rapid,and intelligent Chinese medicine material recommendation methods,which are very important for the inheritance,innovation and promotion of traditional Chinese medicine.Deep learning is a very popular method for artificial intelligence at present,and has achieved gratifying results in many fields such as computer vision,target detection,and natural language processing.This paper uses deep learning methods to achieve feature extraction of tongue coating images and recommended Chinese medicine materials based on the characteristics of tongue coating through deep neural network models.This thesis aims to solve the current recommendation of Chinese herbal medicines that requires specific professionals to use TCM materials.It uses deep neural network models and a large amount of tongue coating image data to extract tongue coating image characteristics and utilizes the strong image between tongue coating images and medicinal materials.Relevance,recommendation of TCM materials and TCM prescriptions that are related to the image of tongue coating have certain auxiliary effects on the recommendation of TCM materials and the generation of TCM prescriptions.Due to the lack of data on tongue coating and prescriptions,there are not enough and high-quality tongue coating image databases.This paper first uses the tongue furrow data acquisition system to directly use smart mobile devices to collect tongue diagnosis images in the hospital outpatient department and collect them.The prescription data is used to construct the tongue fur image database and construct related experimental data sets.Secondly,principal component analysis and Auto-Encoder methods are used to implement the task of extracting tongue coating features.Then a Chinese herbal medicine recommendation based on deep neural network is designed.A model and an algorithm for screening prescriptions from related herbal medicines were selected from the Chinese herbal medicine candidate set.Finally,matrix decomposition and Chinese medicine material recommendation models were respectively implemented,and used for the recommendation of TCM materials and the generation of TCM prescriptions.Thus,Chinese medicine materials and prescriptions for traditional Chinese medicine are recommended based on the characteristics of the tongue image.