基于随机森林算法的中药处方升降浮沉药性预测研究*

作者：郭梦蕊1，陈红梅2，郦春锦3，张 峰3，孙茜茜3，翟华强1，4

单位：1.北京中医药大学中药学院，北京 102400； 2.杭州市中医院，浙江 杭州 310000； 3.杭州唐古信息科技有限公司，浙江 杭州 310000； 4.北京中医药大学中药调剂标准化研究中心，北京 100029

引用：引用：郭梦蕊，陈红梅，郦春锦，张峰，孙茜茜，翟华强.基于随机森林算法的中药处方升降浮沉药性预测研究[J].中医药导报,2025,31(12):279-283.

DOI：10.13862/j.cn43-1446/r.2025.12.044

PDF： 下载PDF

摘要：

目的：探索随机森林算法在中药处方升降浮沉药性预测中的应用，提升处方分析的准确性，为中药处方审核提供科学依据。方法：以《国医大师颜正华临证用药集萃》为数据来源，使用Microsoft Excel 2019构建医案处方数据库，规范治法与饮片名称并确定处方涉及的中药饮片的升降浮沉药性，以治法为依据标记处方的升降浮沉趋势，结合随机森林算法，根据规范后的处方构建升降浮沉药性识别模型并进行预测。结果：共纳入411份中药处方，其中升浮方59份，沉降方182份，升降并用方170份。涉及255味中药饮片，趋向升浮、沉降和双重趋向的饮片分别为50种、155种、37种，另有13种饮片因未被2020年版《中华人民共和国药典》收载，其趋向属性暂未明确。模型在以处方组成、核心药物和剂量作为变量进行训练时，处方药性预测准确率最高。结论：随机森林模型在处方升降浮沉药性预测上具有较高的准确率和稳定性，能够有效识别和预测处方的升降浮沉药性，可初步辅助药师进行中药处方药性审核。

关键词：中药处方；随机森林算法；升降浮沉药性；模型；处方分析

Abstract：

Objective: To explore the application of the random forest algorithm in predicting the ascending, descending, floating and sinking properties of traditional Chinese medicine (TCM) prescriptions, improve the accuracy of prescription analysis, and provide a scientific basis for TCM prescription review. Methods: Using The Collection of Clinical Medicinal Prescriptions by TCM Master Physician YAN Zhenghua as the data source, a database of medical case and prescriptions was constructed with Microsoft Excel 2019. Treatment methods and names of medicinal pieces were standardized, and the ascending, descending, floating and sinking properties of TCM pieces involved in the prescriptions were determined. The ascending, descending, floating and sinking trends of the prescriptions were labeled based on the treatment methods. Combined with the random forest algorithm, an identification model for ascending, descending, floating and sinking properties was constructed and used for prediction based on the standardized prescriptions. Results: A total of 411 TCM prescriptions were included, including 59 ascending-floating prescriptions, 182 descending-sinking prescriptions, and 170 prescriptions with both ascending-floating and descending-sinking properties. A total of 255 types of TCM pieces were involved, among which 50 types had ascending-floating tendency, 155 types had descending-sinking tendency, and 37 types had dual tendencies. In addition, the tendency attributes of 13 types of TCM pieces were not yet clear because they were not included in the Pharmacopoeia of the People's Republic of China (2020 Edition). When the model was trained with prescription composition, core medicines and dosage as variables, the prediction accuracy of prescription properties was the highest. Conclusion: The random forest model has high accuracy and stability in predicting the ascending, descending, floating and sinking properties of TCM prescriptions. It can effectively identify and predict the ascending, descending, floating and sinking properties of prescriptions, and can initially assist pharmacists in the property review of TCM prescriptions.

Key words：traditional Chinese medicine prescriptions; random forest algorithm; ascending, descending, floating and sinking properties; model; prescription analysis

发布时间：2025-12-31

点击量：3