基于改进遗传算法的2型糖尿病中医药有效处方推荐方法研究  

Study on Effective Prescription Recommendation Method of Traditional Chinese Medicine for Type 2 Diabetes Based on Improved Genetic Algorithm

作　　者：刘红萍 杨杰 庞国明[3] 李鹏辉 邢颖 吴敏 温宵宵 李洪皎 文天才[1,2] LIU Hongping;YANG Jie;PANG Guoming;LI Penghui;XING Ying;WU Min;WEN Xiaoxiao;LI Hongjiao;WEN Tiancai(Institute of Information on Traditional Chinese Medicine,China Academy of Chinese Medical Sciences,Beijing 100700,China;Traditional Chinese Medicine Data Center,China Academy of Chinese Medical Sciences,Beijing 100700,China;Kaifeng Traditional Chinese Medicine Hospital,Kaifeng 475000,Henan,China;Institute of Basic Research in Clinical Medicine,China Academy of Chinese Medical Sciences,Beijing 100700,China)

机构地区：[1]中国中医科学院中医药信息研究所,北京100700 [2]中国中医科学院中医药数据中心,北京100700 [3]开封市中医院,河南开封475000 [4]中国中医科学院中医临床基础医学研究所,北京100700

出　　处：《中华中医药学刊》2025年第1期39-43,I0009,共6页Chinese Archives of Traditional Chinese Medicine

基　　金：国家自然科学基金面上项目(82374624);北京市自然科学基金面上项目(7232306);中国中医科学院自主选题项目(ZZ16-XRZ-107-SJ)。

摘　　要：目的改进遗传算法创新推荐度模型,以2型糖尿病(T2DM)电子病历数据为基础进行核心处方挖掘及有效处方推荐。方法基于真实世界电子病历数据,构建T2DM患者有效处方集及不同证型原始处方集数据库;对遗传算法进行改进,优化适应度函数的构建,使提取出的核心有效处方朝着有效处方集的方向进化;基于核心有效处方与原始处方集的相似度关系,创新推荐度模型,通过遍历不同证型原始处方集进行处方推荐度挖掘。结果共使用有效诊疗标准的处方17712条,提取的核心有效处方中包含中药37种;最终挖掘出满足“推荐度≥85%”的处方26条,最大推荐度为97.26%。结论研究改进遗传算法提高了对中药处方集的特征提取和全局搜索能力,并通过提出新的推荐度模型进行临床处方决策推荐,提高了电子病历数据的利用率。Objective To improve the genetic algorithm and innovate the recommendation model,mine the core prescriptions and recommend the effective prescriptions based on the electronic medical records data of type 2 diabetes mellitus(T2DM).Methods Based on the real-world electronic medical records data,the database of effective prescriptions of T2DM patients and original prescriptions of different syndromes was constructed.The genetic algorithm was improved and the construction of fitness function was optimized,so the extracted core effective prescriptions can evolve towards the collection of effective prescriptions.Based on the similarity relationship between the core effective prescriptions and the collection of original prescriptions,the recommendation model was innovated,and the prescription recommendation was mined by traversing the collection oforiginal prescriptions of different syndromes.Results A total of 17712 prescriptions with effective diagnosis and treatment standards were used,and 37 kinds of traditional Chinese medicines were extracted.Finally,26 prescriptions with“recommendation value≥85%”were excavated,and the maximal recommendation value was 97.26%.Conclusion In this study,the improved genetic algorithm has improved the ability of feature extraction and global search for prescription sets of traditional Chinese medicine,and proposed a new recommendation model for clinical prescriptions decision-making,thus improving the utilization rate of electronic medical records data.

关 键 词：遗传算法 核心处方 处方推荐 真实世界数据 2型糖尿病 

分 类 号：R259.871[医药卫生—中西医结合]