关联规则在栀子豉汤类方数据挖掘中的应用

摘要

Objective To discuss the specific role of association rules in the compatibility of herbs and the relationship between herbs and symptoms by analyzing ancient analogous prescriptions Zhizichi Decoction with SQL Server software.Methods The prescriptions which contain Gardenia and Semen Sojae Preparatum were searched in the database of “Ancient Prescriptions for Epidemic Disease”, and they were imported to software with the standardized data of their symptoms and ingredients, then the proper association rules were found. The laws between parameter setting and rule generation were searched through parameter adjustment.Results Many laws of the drug compatibility and relationship between herbs and symptoms were obtained effectively in data mining by adjusting the parameters:The support setting was good for finding common used drug compatibility, and the confidence was good for finding specific drug compatibility and relationship between herbs and symptoms;the item set which had high frequency, sometimes made the rule’s importance low.Conclusion The laws of compatibility of herbs and the relationship between herbs and symptoms for data mining of analogous prescriptions are discovered by analyzing the support, confidence and importance of the association-rules and the clinical magnificence between symptoms and herbs.%目的：采用SQL Server软件对古代治疗疫病方剂中的栀子豉汤类方进行关联分析，探讨关联规则参数设置在类方药物配伍、药-症关联分析中的作用。方法在“古方疫病数据库”中检索包含栀子、淡豆豉的方剂，并将标准化后的症状、药物数据导入SQL Server 2005软件形成关联规则，并通过调整参数，寻找参数设置与产生规则之间的规律。结果通过调整参数可有效挖掘多种不同的药物间和药-症间规则：支持度的设置有助发现常用配伍，置信度的调整有助挖掘特异性的药物、药-症规则；支持度高的项集所在的规则，其重要性反而降低。结论综合关联规则的支持度、置信度、重要性，并结合药、症的实际意义，可对类方中药物配伍和药-症关联开展有效挖掘。