目的深入探索方剂的制方要素，总结方药证治规律，找到对其中药理起关键作用的核心药/药组，为方剂配伍理论研究、药性理论研究、中医临床用药研究提供新方法和现代技术手段。方法从《中国方剂数据库》和《方剂现代应用数据库》中获取治疗中风病的方药文本数据源，经预处理、信息抽取后构建方药矩阵，引入双向聚类方法对方-药两方面数据同时聚类。结果预处理后得到有效记录648首，共计6913味药，生成了100×648“方-药”矩阵，最后得到4组聚类指标，高描述度代表方共8首，总结出针对不同证型配伍用药特点及各证型辨证论治的治则治法及各型常用药对药组。结论本方法适用于中医方剂数据库中方-药知识发现，可有效抽取所需方药信息，梳理出方药证治规律。%Objective To explore the key elements in prescription and summarize the law of prescription-herb treatment, to find core herbs/herb group which play a key role in pharmacological effects, then provide new methods and modern techniques for theory of prescription development and TCM clinical research. Methods The text data sources of stroke disease were retrieved from databases of traditonal Chinese prescription. After pre-processing and information extraction, a prescription-herb matrix was constructed, and then biclustering method was used to simultaneous clustering two dimensions of data. Results Totally 648 effective records were gained after preprocessing, and 6913 herbs were extracted, then a 100×648 prescription-herb matrix was generated. Finally 4-way clustering results were output, along with 8 high descriptive prescriptions, then characteristics of prescription laws of diagnosis and treatment according different syndrome and various types of herbs commonly used were summarized. Conclusion This method can be well applied for prescription-herb knowledge discovery in TCM prescription databases. The prescription-herb information can be effectively extracted, and laws of prescription-herb and syndrome-treament can be mined.
展开▼
