基于频繁闭项集的新联系分类算法accf.pdfVIP

第41 卷 第1 期 电 子 科 技 大 学 学 报 Vol. 41 No. 1 20 12年1月 Journal of University of Electronic Science and Technology of China Jan. 20 12 基于频繁闭项集的新关联分类算法ACCF 1 1 2 1 李学明 ，杨 阳 ，秦东霞，周尚波 (1. 重庆大学计算机学院 重庆 沙坪坝区 400044; 2. 周口师范学院计算机科学与技术系 河南 周口 466200) 【摘要】提出了基于频繁闭项集的新关联分类算法ACCF 。ACCF 首先挖掘出所有频繁闭项集(CFIs)和候选分类关联规则， 然后从候选分类关联规则中产生和筛选出若干规则，并用其构建分类器；在分类应用时，采用了一种新的匹配方式对分类实 例进行分类。通过理论分析和对18个UCI公共数据集的实验结果表明 ，ACCF 不仅能挖掘出高质量且不丢失信息的关联分类规 则，而且大大减少了关联分类规则的数量，在分类准确率上也比现有的关联分类典型算法更高。 关 键 词 关联分类; 类关联规则; 频繁闭项集; 数据挖掘 中图分类号 TP 312 文献标识码 A doi:10.3969/j.issn.1001-0548.2012.01.020 Associative Classification Based on Closed Frequent Itemsets 1 1 2 1 LI Xue-ming , YANG Yang , QIN Dong-xia , and ZHOU Shang-bo (1. College of Computer Science，Chongqing University Shapingba Chongqing 400044; 2. Department of Computer Science and Engineer, Zhoukou Normal University Zhoukou Henan 466200) Abstract A new associative classification method named ACCF is presented based on the closed frequent itemsets. This method first mines all closed frequent itemsets and the candidate class association rules (CARs), and then constructs classifier based the selected CARs from the candidate CARS. The new instances are finally classified by a new way. Our theoretical analysis and substantial experiments on 18 datasets from UCI repository of machine learning databases show that ACCF is highly effective at classification of various kinds of datasets. Compared with the typical associative classification algorithms, ACCF can mine much less CARs and has higher a