融合多维度相似算法的中药复方个性化煎煮推荐研究

Personalized Decoction Recommendation of Traditional Chinese Medicine Compound Formulas with Algorithm of the Multi-dimensional Similarities

目的:为实现中药汤剂智能煎煮设备煎煮参数快速推荐,提出一种融合方剂多维度相似算法,借鉴古代经典名方的个性化煎煮推荐。方法:选取30首经典名方,给出算法推荐的煎煮参数值,比较与典籍记载对应的煎煮参数值之间的差异。针对中药复方的组成成分、剂量以及功用维度特性,分别采用Jaccard相似系数、夹角余弦相似度和潜在狄利克雷分布(LDA)主题模型相似度算法,计算实验方剂与煎煮数据库中方剂间的相似度,将各维度最为相似方剂的煎煮参数加权融合,获取30首经典名方的个性化煎煮参数。结果:加水量、武火及文火时长参数值之间的相对误差分别为5.3%、3.4%、7.1%。结论:2组煎煮参数值较为接近,较好地借鉴和传承经典名方煎煮工艺。

Objective:To achieve rapid recommendation of decoction parameters for intelligent decoction equipment of traditional Chinese medicine(TCM)decoction,a multi-dimensional similarity algorithm integrating classical formulas was proposed to provide personalized decoction recommendations by drawing on ancient classical prescriptions.Methods:Thirty classical prescriptions were selected,and the decoction parameters recommended by the algorithm were compared with those recorded in ancient texts.The Jaccard similarity coefficient,cosine similarity,and Latent Dirichlet Allocation(LDA)topic model similarity algorithms were employed to calculate the similarity between experimental formulas and those in the decoction database,based on the composition,dosage,and functional characteristics of TCM compound formulas.The decoction parameters of the most similar formulas in each dimension were weighted and integrated to obtain personalized decoction parameters for the thirty classical prescriptions.Results:The relative errors in water volume,strong fire duration,and mild fire duration were 5.3%,3.4%,and 7.1%,respectively.Conclusion:The two sets of decoction parameters were relatively close,effectively drawing on and inheriting the decoction techniques of classical prescriptions.