Application of Latent Dirichlet Allocation model based on attribute knowledge embedding in traditional Chinese medicine recommendation
Objective:A data-driven medicinal materials recommendation method helps traditional Chinese medicine(TCM)physicians to make scientific treatment prescriptions more accurately and intelligently in real clinical practice,and can also provide a scientific basis for the development of TCM diagnosis and treatment.Methods:24 127 TCM prescription records were analyzed by text mining method,and the process of generating prescriptions was simulated based on TCM theory,and the knowledge of the rich information of symptoms and medicinal materials and their interrelationships was taken into account.A TCM recommendation method based on Latent Dirichlet Allocation(LDA)topic model embedded with attribute knowledge network was proposed,and the attribute knowledge network of TCM contains rich attribute information of medicinal materials and pharmacological effects,which enhanced the topic model.Results:The experimental results showed that the prediction perplexity,accuracy and average Area Under Curve(AUC)of the model performed better compared with those of the baseline topic model under the optimal embedding coefficients.Conclusion:The proposed method is beneficial in improving model stability and accuracy of medicinal materials recommendation,and better undertaks tasks such as diagnosis and treatment pattern mining.
traditional Chinese medicinedata miningtopic modelattribute knowledge networkherb recommendation
NETL
NSTL
万方数据
