<p>Multi-label feature selection technology shows significant effectiveness when dealing with high-dimensional data and thus has received extensive research attention. However, most of the existing methods assume that there is a linear relationship between features and labels. This assumption is usually not valid in practical situations, especially in complex high-dimensional spaces. In addition, although graph structure-based methods are commonly used to model the relationship between labels and features, they usually focus on local smoothness and ignore global structural information, thus limiting the ability to capture global relationships. In order to address these challenges, this paper proposes a multi-label feature selection algorithm based on logistic regression and random walk strategy, which firstly utilizes logistic regression to establish probabilistic relationships between instances and labels, effectively capturing nonlinear dependencies that may be ignored by linear models. Secondly, a random walk strategy is introduced on the feature graph to enable the model to consider both local feature similarities and global structural dependencies. Thirdly, label manifold learning is combined with logistic regression to further explore the intrinsic associations among labels. Finally, a sparser <InlineEquation ID="IEq1"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="10115_2025_2532_Article_IEq1.gif" Format="GIF" Height="19" Rendition="HTML" Resolution="72" Type="Linedraw" Width="22" /> </InlineMediaObject> <EquationSource Format="TEX">\(l_{2,0}\)</EquationSource> <EquationSource Format="MATHML"><math> <msub> <mi>l</mi> <mrow> <mn>2</mn> <mo>,</mo> <mn>0</mn> </mrow> </msub> </math></EquationSource> </InlineEquation>-norm is introduced for better feature selection, and an efficient coordinate descent method is used to optimize the model efficiently. Extensive experimental results on 15 multi-label datasets fully show the excellent performance of the proposed algorithm.</p>

错误:搜索内容不能为空,请输入英文关键词
错误:关键词超出字数限制,请精简
高级检索

Multi-label feature selection based on logistic regression and random walk strategy

  • Guanghui Liu,
  • Qiaoyan Li,
  • Xiaofei Yang,
  • Zhiwei Xing,
  • Yingcang Ma

摘要

Multi-label feature selection technology shows significant effectiveness when dealing with high-dimensional data and thus has received extensive research attention. However, most of the existing methods assume that there is a linear relationship between features and labels. This assumption is usually not valid in practical situations, especially in complex high-dimensional spaces. In addition, although graph structure-based methods are commonly used to model the relationship between labels and features, they usually focus on local smoothness and ignore global structural information, thus limiting the ability to capture global relationships. In order to address these challenges, this paper proposes a multi-label feature selection algorithm based on logistic regression and random walk strategy, which firstly utilizes logistic regression to establish probabilistic relationships between instances and labels, effectively capturing nonlinear dependencies that may be ignored by linear models. Secondly, a random walk strategy is introduced on the feature graph to enable the model to consider both local feature similarities and global structural dependencies. Thirdly, label manifold learning is combined with logistic regression to further explore the intrinsic associations among labels. Finally, a sparser \(l_{2,0}\) l 2 , 0 -norm is introduced for better feature selection, and an efficient coordinate descent method is used to optimize the model efficiently. Extensive experimental results on 15 multi-label datasets fully show the excellent performance of the proposed algorithm.