Patient-Oriented Chinese Patent Drug Recommendation System

With the development of Internet e-commerce and intelligent medical-assisted diagnosis technology,more and more patients purchase medicines that suit their symptoms through e-commerce platforms.When people buy medicines online,they often face the problem of blind selection and no professional guidance.According to the big data analysis of medication errors in medical institutions in 19 provinces in 2017,60.95%of medication errors came from doctors and 34.57%came from pharmacists.In the actual diagnosis and treatment process,if the treatment time is delayed or the body is irreversibly injured due to medication errors,it will cause huge losses to individuals and the entire family.At this stage,the drug recommendation system is mainly for doctors,which can assist doctors in making quick decisions,thereby improving their work efficiency.However,for the Internet,e-commerce platforms,and regions with relatively tight medical resources,a patient-oriented Chinese patent drug recommendation system is required to improve the accuracy of diagnosis and treatment effects.Aiming at the problem of low prediction accuracy in common drug recommendation algorithms and the need to construct a large number of user drug scoring vectors,to achieve patient-oriented Chinese patent drug recommendation and improve the accuracy of a recommended medication.Fully investigated the deep learning text classification model,recommendation system ranking model,especially the application of related models in TCM diagnostics,and researched TCM syndrome recall classification and TCM syndrome sorting algorithms on this basis.The main research work and innovations of this paper are as follows:(1)A TCM syndrome recall model based on a text convolutional neural network is proposed.The main design idea of the model refers to the deep pyramid text convolution structure.To improve the recall accuracy of the model,a deep pyramid text convolution neural network with embedded patient features is constructed.The model uses the embedding layer to fuse the patient’s symptom features and basic information features,uses a multi-scale convolution kernel to extract the basic features of the text area,and then uses continuous equal-length convolution blocks to complete the long-distance association of the text information and repeats the down-sampling operation And the shortcut connection extracts complex context information features,which improves the recall accuracy of the TCM syndrome candidate set.(2)A TCM syndrome recommendation ranking model based on deep learning is proposed to strengthen the use of patient cross-feature information.The model is based on DeepFM,combined with the patient’s current characteristics and the candidate syndrome types after recall for sorting prediction.The factorization layer and the hidden layer process the low-and high-level features separately,and combine the outputs of the two to complete the fusion of low-level features and high-level features.Deepen the network depth through residual connection and accelerate the training of the model.Finally,find out the corresponding drugs according to the predicted TCM syndrome type,and complete the ranking recommendation of the TCM recommendation system.(3)Analyze the function and performance requirements of the patient-oriented Chinese patent drug recommendation system,adopt high independence,low coupling,and other design concepts,design and implement a patient-oriented Chinese patent medicine recommendation system,which can help patients find the best in the massive information.Symptomatic medicine.Through a comprehensive analysis of the system’s computing efficiency,platform performance,and visualization framework,it is determined that Flask is the back-end framework,JavaScript is the front-end language,Echarts is the visualization component,and MySQL is the main technical route of the database.Through PyTorch,the construction of the TCM syndrome classification model and the ranking model was completed,and on this basis,a patient-oriented TCM recommendation system was built.The actual case was tested,and the results showed that the performance of the patient-oriented TCM recommendation model meets the expected effect of recommendation.