一种基于关联规则的多类标分类算法  

Multi-label classification algorithm based on association rules

作　　者：李宏[1] 李博[1] 吴敏[1] 陈松乔[1] 

机构地区：[1]中南大学信息科学与工程学院,长沙410083

出　　处：《控制与决策》2009年第4期574-578,582,共6页Control and Decision

基　　金：国家杰出青年科学基金项目(60425310);中南大学博士后基金项目(2008)

摘　　要：提出了一种基于关联规则的多类标算法(MLAC),利用多类标FP-tree来分解组合生成多类标规则,并通过组合多重关联规则分类器进行分类预测,降低了由高维属性带来的高计算复杂度,有效地提高了算法的性能和效率.针对多类标数据集的实验结果表明,MLAC算法在性能和效率等方面均优于ML-KNN等多类标分类算法.A multi-label classification algorithm based on association rules （MLAC） is proposed, which uses multilabel FP_tree to resolve multi-label data and merger labels with the same attributes to generate multi-label rules. The classification prediction is conducted by assembling the classifiers. So the computational complexity brought by high dimensional attribute decreases while the performance and efficiency increases. The results of subsequent experiment based on multi-label data show that MLAC achieves better performance and efficiency than ML-KNN and other multilabel classification algorithms.

关 键 词：分类 关联规则 多类标 单类标 组合 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]