目的 探讨汉代以来著名医家治疗痰饮的方剂用药与组方规律，以期对痰饮临床治疗及中药新药研发提供参考。方法 收集整理汉代以来著名医家的痰饮论述与医案共涉及208首方剂，录入中医传承辅助系统（V2.5），采用改进的互信息法、关联规则、复杂系统熵聚类、无监督的熵层次聚类，进行痰饮数据库单药、对药、角药的统计、关联规则分析、提取核心组合与新方发现，进行数据挖掘研究。结果 最常见单药是半夏（117次，占56.25%）；最常见对药是半夏-茯苓（62次，占29.81%）；最常见角药是半夏-陈皮-茯苓（33次，占15.87%）；关联度最高的药物组合是生姜 + 茯苓→半夏（置信度为0.8636）；得到18组3味药的核心组合，核心药物为：半夏、茯苓、陈皮、生姜、甘草、桂枝等；得到治疗痰饮9个新候选处方。结论 提示今后痰饮治疗及新药研发时，单药、对药、角药优先考虑：半夏、陈皮、茯苓；配伍优先考虑：生姜、茯苓配伍半夏；核心方剂优先考虑：杏仁-麻黄-桑白皮；新候选处方优先考虑：杏仁、麻黄、桑白皮、阿胶、款冬花。

Objective To investigate famous ancient traditional Chinese medicine (TCM) doctors prescribing herbal medicines for treatment of phlegm-retained fluid since the Han Dynasty, so as to provide basis for clinical treatment on phlegm-retained fluid as well as for researching and developing new medicines for treating phlegm-retained fluid.Methods A total of 208 TCM prescriptions were collected and input into TCM Inheritance Auxiliary Plat form Version 2.5. Using the method of improved mutual information entropy clustering, association rules, complex systems entropy clustering and unsupervised entropy hierarchical clustering, the single drug, pairs of drugs, triangle drugs, association rules of drugs, evolution and extraction of combination of the core drugs, and prescriptions newfound were analyzed.Results The most commonly used herbal medicine was Banxia (117, 56.25%), and the most common pair of drugs was Banxia-Fuling (62, 29.81%), the most common triangle drugs was Banxia-Chenpi-Fuling (33, 15.87%). The drugs combination with high correlation was Shengjiang-Fuling→Banxia, with confidence of 0.863 6. Altogether 18 sets of triangle drugs were extracted, in which the core drugs were Banxia, Fuling, Chenpi, Shengjiang, Gancao, Guizhi. And 9 new prescriptions were available.Conclusion For research and development of new drugs on treating phlegm-retained fluid, Banxia, Chenpi and Fuling should be considered first; prior compatibility rules are Shengjiang, Fuling with Banxia; Xingren, Mahuang and Sangbaipi should be the core formula drugs; when formulate a new prescription, Xingren, Mahuang, Sangbaipi, E jiao and Kuandonghua deserve the priority.