价格适应性的药品关联规则学习及推荐  

Price adaptive medicine association rule learning and medicine recommendation

作　　者：刘卓名 陈干 陈碧毅 黄玲[1] 王昌栋[1] LIU Zhuoming;CHEN Gan;CHEN Biyi;HUANG Ling;WANG Changdong(School of Data and Computer Science,Sun Yat-Sen University,Guangzhou Guangdong 510006,China)

机构地区：[1]中山大学数据科学与计算机学院,广州510006

出　　处：《计算机应用》2020年第S01期80-84,共5页journal of Computer Applications

基　　金：中山大学高校基本科研业务费—新兴学科交叉学科资助计划项目(19lgjc10);国家自然科学基金面上项目(61876193);广东省自然科学基金杰出青年科学基金资助项目(2016A030306014)。

摘　　要：传统的关联规则(AR)学习算法根据规则出现的频率为每一个关联规则左端项(LEI)找到在频率意义下最优的右端项(REI)进行推荐。现实生活中影响关联规则重要程度的因素很多,而传统的学习算法仅考虑了频率,因此不能给出个性化的结果;此外,传统算法也没有考虑关联规则右端项之间的关系,无法生成一组左端项相同而右端项类似并且可以相互替代的关联规则,因此通过其他因子对规则进行筛选时会失去一些重要信息。针对这些问题,提出一种价格适应性的药品关联规则学习及推荐算法,考虑子关联规则之间以及子母关联规则之间的关系,计算关联规则右端项之间的互信息,确定了频繁项集置信度的阈值。实验结果表明,利用MIMIC数据库,所提算法挖掘出了数据中左端项相同且右端项为药效相似但价格不同的一系列关联规则,并能根据给定左端项和价格阈值完成关联规则的推荐任务。Conventional Association Rule(AR)learning methods discover Right-End-Item(REI)for a specific LeftEnd-Item(LEI)of the ARs according to the rules’ frequency in the database. However,these methods may not provide personalized results,because there are many other ignored potential factors,affecting the importance of ARs in real life.Besides,these methods do not consider the relationship of the REIs of the ARs and cannot find a group of ARs that have the same LEI but similar REIs which can be replaced with others,which leads to the information loss when screening the ARs by using some factors. To address these issues,by proposing the price adaptive medicine association rule learning and medicine recommendation algorithm,considering the relationship between the sub-ARs and sup-ARs,sub-ARs and sub-ARs and calculating the mutual information between the REIs,the thresholds of confidence in frequent itemset were determined. By making use of the prescriptions in the MIMIC Ⅲ,the proposed algorithm finds some useful ARs,which have the same LEI and different REIs with similar medical effect and different prices,and makes use of the given LEI and a price threshold to recommend medicine.

关 键 词：关联规则 价格适应性 药品推荐 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]