目的 评价宋元时期消渴用药规律及证治特点.方法 计算机检索《中华医典》宋元时期综合类方书的消渴证治方剂,采用隐结构结合综合聚类模型,构建消渴证治用药规律的隐树结构图,结合隐类概率、条件概率、互信息、信息覆盖率等用药量化数据,人工判读诠释各隐类消渴证治用药规律特征.结果 宋元时期消渴证治用药分为11个隐类,分别以天花粉与黄连、熟地与黄芪、肉苁蓉与山茱萸、人参与当归、桑白皮与赤茯苓、知母与麦门冬、泽泻与石脂、鸡内金与生地、五味子与磁石、牡蛎与苦参、龙骨与桑螵蛸为核心代表药物.通过以药测证,宋元时期消渴主要表现为阴虚阳浮证、阴虚毒炽证、阴虚火旺证、阴虚不固证、肝肾阴虚证、气血两虚证、气阴两虚证、湿热血瘀证、水饮内停证、湿浊内阻证等.结论 隐结构模型与综合聚类模型结合方法,较适宜于疾病证治用药规律研究,亦可用于辅助验证疾病证候要素研究结果.“,”Objective To evaluate medication laws and syndrome-treatment features of diabetes syndrome in Song and Yuan Dynasties.Methods Recipes of diabetes syndrome-treatment in Song and Yuan Dynasties collected in Chinese Medical Classics were retrieved by computer.Using latent structure model and comprehensive clustering model,latent tree structure chart of regular laws for diabetes syndrome-treatment was constructed.Data were quantified by combining latent probability,conditional probability,mutual information,and cumulative information coverage rate,and regular laws for diabetes syndrome-treatment were interpreted manually.Results There were 11 latent kinds for diabetes syndrometreatment,represented as Radix Trichosanthis and Coptis Chiinensis,Rehmannia and Astragalus Membranaceus,Citanche Deserticola and Cornus Officinalis,Ginseng and Angelica sinensis,Cortex Mori and Poria Rubra,Anemarrhena Asphodeloides and Radix Ophiopogonis,Alisma Orientale and Halloysitum Rubrum,Endothelium Corneum Gigeriae Galli and Rehmannia glutinosa,Schisandra Chinensis and Magnetium Magnetite,Concha Ostreae and Sophora Flavescens,Os Draconis and Ootheca Mantidis.By herbs based syndromes testing,main syndromes of diabetes syndrome in Song and Yuan Dynasties were yin deficiency yang floating syndrome,yin deficiency toxic heat syndrome,yin deficiency induced fire hyperactivity syndrome,yin deficiency induced unconsolidation syndrome,Gan-Shen yin deficiency syndrome,qi-blood deficiency syndrome,qi-yin deficiency syndrome,damp heat blood stasis syndrome,internal retention of water syndrome,damp turbidity internal resistance syndrome,and so on.Conclusions Latent structure model combined with comprehensive clustering model were more suitable for syndrome-treatment laws.It could also be used for auxiliarily verifying results of syndrome factors.