【摘要】 传统的数据挖掘方法一般从组方中所有的药物出发，挖掘药物的用药规律，计算量大，且仅仅基于药物频次对组方进行研究，忽略了药物剂量因素，难以发现频次低但剂量占比高的具有良好疗效的药物。针对以上问题，提出一种改进的基于效用度(Effect Degree,ED)核心药物发现算法，并将基于效用度的点式互信息(Pointwise Mutual Information with Herb Pair ED,PMIED)与节点度结合，定义一种新的加权相关系数作为药物权重，在所发现的核心药物中运用层次聚类算法研究用药规律。实验结果表明，该算法可有效挖掘出组方中的核心药物，经过分析，所发现的核心药物和药物组合均对痰瘀互阻证具有良好疗效。更多还原

【Abstract】 The traditional data mining method generally starts from all the drugs in the prescriptions to mine the medication rule of drugs,which is computationally intensive and only researches the prescriptions based on the frequency of drugs,ignoring the drug dosage factor.It is difficult to find drugs with low frequency but high dosage ratio and good efficacy.To solve the above problems,this paper proposes an improved core drug discovery algorithm based on Effect Degree (ED),and combines Pointwise Mutual Information with Herb Pair ED (PMIED) with degree of a node.It defines a new weighted correlation coefficient as drug weight,and uses Hierarchical Clustering algorithm to research the medication rule among the discovered core drugs.The experimental results show that the algorithm can effectively mine the core drugs in the prescriptions.After analysis,the discovered core drugs and drug combinations have a good effect on the phlegm-blood stasis syndrome.更多还原