基于正负类边界距离的多标签数据属性约简
Attribute Reduction for Multi-Label Data Based on Boundary Distance between Positive and Negative Classes
摘要: 本文首先将多标签分类问题分解为一系列单标签二分类子问题，每一个子问题对应一个标签。子问题的正类定义为具有该标签的样本，负类定义为不具有该标签的样本。在给定的属性子集下，计算出正负类样本之间的最小距离，即分类边界的最小距离。将子问题分类边界最小距离求和定义为依赖度函数，并将此依赖度函数作为属性子集重要度评价指标。然后建立了所提出依赖度函数关于属性子集的单调性，并通过最大化依赖度函数给出了属性约简的定义。最后，设计了一种基于正负类边界距离的属性约简算法，并在实际的多标签数据集上进行了实验，实验结果表明，所提约简算法能够建立合理的属性重要度排序，有效地去除冗余属性。
Abstract: In this paper, the multi-label classification problem is decomposed into a series of single-label bi-nary classification sub-problems, each of which corresponds to one label. The positive class of the sub-problem is defined as the sample with the label, and the negative class is defined as the sample without the label. Under the given attribute subset, the minimum distance between positive and negative class samples is calculated, that is, the minimum distance of classification boundary. The sum of the boundary distances for all sub-problems is defined as dependency function, which is used as the evaluation of the importance of the attribute subset. The monotonicity of the proposed dependence function with respect to the attribute subset is established, and the definition of the attribute reduction is given by maximizing the dependence function. Finally, an attribute reduction algorithm based on positive and negative class boundary distance is designed, and the experiments are carried out on actual multi-label data sets. The experimental results show that the proposed algorithm can establish a reasonable ranking of attribute importance and effectively remove redundant attributes.
文章引用：纪思南. 基于正负类边界距离的多标签数据属性约简[J]. 计算机科学与应用, 2020, 10(9): 1549-1558. https://doi.org/10.12677/CSA.2020.109163

参考文献

[1] Zhang, M.L. and Zhou, Z.H. (2014) A Review on Multi-Label Learning Algorithms. IEEE Transactions on Knowledge and Data Engineering, 26, 1819-1837. [Google Scholar] [CrossRef] 
[2] Xu, S.P., Yang, X.B., Yu, H.L., Yu, D.J., Yang, J.Y., Eric, C.C. and Tsang, J. (2016) Multi-Label Learning with Label-Specific Feature Reduction. Knowledge-Based Systems, 104, 52-61. [Google Scholar] [CrossRef] 
[3] Zhu, P.F., Xu, Q., Hu, Q.H., Zhang, C.Q. and Zhao, H. (2018) Multi-Label Feature Selection with Missing Labels. Pattern Recognition, 74, 488-502. [Google Scholar] [CrossRef] 
[4] Li, H., Li, D.Y., Zhai, Y.H., Wang, S.G. and Zhang, J. (2016) A Novel Attribute Reduction Approach for Multi-Label Data Based on Rough Set Theory. Information Sciences, 367-368, 827-847. [Google Scholar] [CrossRef] 
[5] Li, Y.W., Lin, Y.J., Liu, J.H., Weng, W., Shi, Z.K. and Wu, S.X. (2018) Feature Selection for Multi-Label Learning Based on Kernelized Fuzzy Rough Sets. Neurocomputing, 318, 271-286. [Google Scholar] [CrossRef] 
[6] Liu, J.H., Lin, Y.J., Li, Y.W., Weng, W. and Wu, S.X. (2018) Online Multi-Label Streaming Feature Selection Based on Neighborhood Rough Set. Pattern Recognition, 84, 273-287. [Google Scholar] [CrossRef] 
[7] Lin, Y.J., Hu, Q.H., Zhang, J. and Wu, X.D. (2016) Mul-ti-Label Feature Selection with Streaming Labels. Information Sciences, 372, 256-275. [Google Scholar] [CrossRef] 
[8] Fan, X.D., Zhao, W.D., Wang, C.Z. and Huang, Y. (2018) Attribute Re-duction Based on Max-Decision Neighborhood Rough Set Model. Knowledge-Based Systems, 151, 16-23. [Google Scholar] [CrossRef] 
[9] Lin, Y.J., Li, Y.W., Wang, C.X. and Chen, J.K. (2018) Attribute Re-duction for Multi-Label Learning with Fuzzy Rough Set. Knowledge-Based Systems, 152, 51-61. [Google Scholar] [CrossRef] 
[10] Wang, C.X., Lin, Y.J. and Liu, J.H. (2019) Feature Selection for Multi-Label Learning with Missing Labels. Applied Intelligence, 49, 3027-3042. [Google Scholar] [CrossRef] 
[11] Liu, J.H., Li, Y.W., Weng, W., Zhang, J., Chen, B.H. and Wu, S.X. (2020) Feature Selection for Multi-Label Learning with Streaming Label. Neurocomputing, 387, 268-278. [Google Scholar] [CrossRef] 
[12] 段洁, 胡清华, 张灵均, 钱宇华, 李德玉. 基于邻域粗糙集的多标记分类特征选择算法[J]. 计算机研究与发展, 2015, 52(1): 56-65.
[13] Lin, Y.J., Hu, Q.H., Liu, J.H. and Duan, J. (2015) Multi-Label Feature Selection Based on Max-Dependency and Min-Redundancy. Neurocomputing, 168, 92-103. [Google Scholar] [CrossRef] 
[14] Fan, X.D., Chen, Q., Qiao, Z.J., Wang, C.Z. and Ten, M.Y. (2020) Attribute Reduction for Multi-Label Classification Based on Labels of Positive Region. Soft Computing, 1-11. [Google Scholar] [CrossRef] 
[15] Zhang, M.L. and Zhou, Z.H. (2006) ML-KNN: A Lazy Learning Approach to Multi-Label Learning. Pattern Recognition, 40, 2038-2048. [Google Scholar] [CrossRef]