基于HPD算法的中药药对挖掘方法

A HPD-based Approach for Discovering the Herb-pairs

人工智能方法辅助中药药对发现对于中医发展具有重要的现实意义,但当前的药对挖掘方法大多基于现有的单一的机器学习模型,使得药对发现的质量不高。本文旨在找出中医药候选药对,通过分析已知药对中饮片之间的特性和耦合关系,提出了一种改进的机器学习算法(Herb-Pairs Discovering,HPD)。该方法采用药对属性相关度评分方法,找出其中不低于给定阈值的饮片对。在基于朴素贝叶斯对饮片作用分类方法基础上利用决策树算法对饮片组合进行分析进而找出预测药对。在实际中医药测试数据上进行实验,结果分析表明,HPD算法得到的药对召回率达到82.7%,准确率达到80.6%,说明HPD算法可以有效地发现中医药饮片集合中潜在的药对。

The artificial intelligence method to assist Chinese medicine has important practical significance for the development of Chinese medicine, but the current herb pairs discovery methods are mostly based on the existing single machine learning model, which makes the quality of herb pairs discovery not high. This paper is proposed for the discovery of Chinese herb pairs by analyzing the characteristics and coupling relationship between known medicine pairs and Chinese decoction pieces, an improved machine learning algorithm(Herb-Pairs Discovering, referred to as HPD).This method uses the scoring method of herb pairs attribute correlation to find the pairs of decoction pieces that is not lower than a given threshold. On the basis of the classification method based on Naive Bayes on the effect of the decoction pieces, the decision tree algorithm is used to analyze the combination of decoction pieces to find the predicted herb pairs. Experiments on actual Chinese medicine test data showed that the herb pairs recall rate obtained by the HPD algorithm reached 82.7%, and the accuracy rate reached 80.6%, indicating that the HPD algorithm can effectively find the potential herb pairs in the collection of Chinese medicine decoction pieces.