摘要
针对多标签数据的标签相关性和高维问题,提出一种基于奇异值分解—偏最小二乘回归的多标签分类算法,该算法可以对多标签数据进行维数约简和回归分析。首先,将类别标签集合作为整体处理,对标签相关性进行考察;其次,利用奇异值分解(SVD)技术得到样本和标签空间的得分向量,实施降维;最后,在偏最小二乘回归(PLSR)的基础上构建多标签分类模型。实验结果表明,在四种维数较高的真实数据集上,该算法可以获得有效的分类结果。
To tackle multi-label data with high dimensionality and label correlations, a multi-label classification approach based on Singular Value Decomposition (SVD)-Partial Least Squares Regression (PLSR) was proposed, which aimed at performing dimensionality reduction and regression analysis. Firstly, the label space was taken into a whole so as to exploit the label correlations. After that, the score vectors of both the instance space and label space were obtained by SVD, which was used for dimensionality reduction. Finally, the model of multi-label classification was established based on PLSR. The experiments performed on four real data sets with higher dimensionality verify the effectiveness of the proposed method.
出处
《计算机应用》
CSCD
北大核心
2014年第7期2058-2060,2089,共4页
journal of Computer Applications
基金
国家自然科学基金资助项目(61100119
61272468)
中国博士后科学基金资助项目(2013M530072)
模式识别国家重点实验室开放基金资助项目(201204214)
浙江省自然科学基金资助项目(LY14F020012)
关键词
多标签分类
奇异值分解
偏最小二乘回归
维数约简
标签相关性
multi-label classification
Singular Value Decomposition (SVD)
Partial Least Squares Regression (PLSR)
dimensionality reduction
label correlation