中医关联规则分析平台存在的缺陷及完善方法  

Loopholes and perfecting measures in traditional Chinese medicine association-rule mining analysis platform

作　　者：李浩[1] 侯贺祥 倪健[1] 唐民科[1] 廖国龙 马星光[1] LI Hao;HOU He-xiang;NI Jian;TANG Min-ke;LIAO Guo-long;MA Xing-guang(Beijing University of Chinese Medicine,Beijing 102488,China)

机构地区：[1]北京中医药大学,北京102488

出　　处：《中华中医药杂志》2022年第11期6671-6674,共4页China Journal of Traditional Chinese Medicine and Pharmacy

基　　金：国家自然科学基金项目(No.81673609);中南中医药防治重大疾病国际合作基地(No.GZYYGJ2021047)。

摘　　要：目的:探讨使用项集与规则提升度指标对完善中医关联规则分析的作用。方法:以《伤寒论》桂枝汤类19首方剂为例,使用R语言环境,基于Aprioir算法进行药物关联规则分析。结果:设置最小支持度为0.2,最小置信度为0.8,未引入提升度,频次最高前3组药物规则为{桂枝}->{炙甘草}、{炙甘草}->{桂枝}、{大枣}->{生姜}。设置提升度>1,频次最高的前3组药物规则为{大枣}->{生姜}、{生姜}->{大枣}、{大枣,炙甘草}->{生姜}。最终获得15个药物核心组合。结论:从集合运算的角度来看,由于桂枝汤类方剂必含桂枝、炙甘草,因此{桂枝}->{炙甘草}类规则并无意义。故中药关联规则分析时应引入项集提升度与规则提升度,以避免无效规则的出现,让关联规则分析平台更好地为中医药服务。Objective: To discuss the role to improve the analysis of traditional Chinese medicine association-rule mining using item lift and rule lift. Methods: Taking 19 prescriptions of Guizhi Decoction in the Treatise on Febrile Diseases as an example. The prescriptions to do association-rule mining of Chinese herbs based on Aprioir algorithm in R was used. Results:Firstly, we restricted the minimum support to 0.2, the minimum confidence to 0.8, and don’t introduce rule lift and item lift.The top 3 association-rules with the highest frequency were {cassia twig}->{radix glycyrrhizae preparata}, {radix glycyrrhizae preparata}->{cassia twig}, {jujube}->{ginger}. Then, we restricted rule lift and item lift greater than 1. The top 3 associationrules with the highest frequency were {jujube}->{ginger}, {ginger}->{jujube}, {jujube, radix glycyrrhizae preparata}->{ginger}.Finally, 15 drug core combinations were obtained. Conclusion: From the point of view of set operations, since Guizhi Decoction must contain cassia twig and radix glycyrrhizae preparata, {cassia twig}->{radix glycyrrhizae preparata} is meaningless.Therefore, in the research of association-rule mining of Chinese herbs based on traditional Chinese medicine association-rule mining analysis platform, we can avoid invalid association-rule by considering item lift and rule lift to make the platform serve traditional Chinese medicine better.

关 键 词：APRIORI算法 关联规则 中药配伍 规则提升度 项集提升度 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论] R2-03[自动化与计算机技术—计算机科学与技术]