融合布尔矩阵和项目特性的关联规则挖掘算法  

Association Rule Mining Algorithms Combining Boolean Matrix and Project Characteristics

作　　者：田建勇 石林江 TIAN Jian-yong;SHI Lin-jiang(School of Electronic and Information Engineering,Anshun University,Anshun 561000,China)

机构地区：[1]安顺学院电子与信息工程学院,贵州安顺561000

出　　处：《控制工程》2020年第6期1004-1011,共8页Control Engineering of China

基　　金：贵州省科技厅、安顺市政府、安顺学院三方联合基金项目(黔科合LH字[2017]7046号)。

摘　　要：针对Apriori算法多次扫描产生大量候选集,挖掘大数据库关联规则效率较低等问题,提出了一种融合布尔矩阵和项目特性的关联规则挖掘算法。算法将事务集看作行向量,项目集看作列向量,通过扫描事务数据集构建反映事务集和项目集关系的布尔矩阵,根据事务数据集布尔矩阵得到关联规则;同时将用户对某一商品的兴趣扩展为对商品特性的兴趣,并根据挖掘出特性间的置信度和用户预测评分对数据项目进行评分,然后采用Top-N推荐算法对关联规则进行推荐。实例分析推演了本算法关联挖掘的流程,仿真实验分析了本算法支持度阈值和置信度阈值与F1-Score值间的关系,确定了最优阈值。对比实验表明,本算法在关联规则推荐中的准确率、召回率和F1-Score值都均高于其他两种同类算法。At present,there are a large number of candidate sets generated during the multiple scanning process of the Apriori algorithm,and the association rules couldn’t be effectively extracted,an association rule mining algorithm that combines Boolean matrices and project characteristics is proposed in this paper.The transaction set was treated as a row vector,the item set was treated as a column vector,and the Boolean matrix that reflects relationship between the transaction set and the item set was built by scanning the transaction data set.At the same time,the user’s interest in a certain product was extended to the interest of the product characteristics,the data items were scored according to the confidence between the extracted features and the user’s predicted score,and then the Top-N recommendation rule was used to recommend the association rules.After the example analysis,the process of association mining of this algorithm was deduced,the simulation experiment analyzed the relationship between the support threshold,the confidence threshold and the F1-Score value,the optimal threshold was determined.The comparison experiments show that the accuracy,recall rate and F1-Score value of the algorithm in the association rule recommendation are higher than the other two similar algorithms.

关 键 词：关联规则 APRIORI算法 布尔矩阵 项目特性 项目评分 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]