Scholarlink

<p>Feature selection is vital in machine learning and data analysis, as it enhances model performance, reduces computational costs, and improves efficiency. However, existing embedded methods, particularly those based on regression, often assume a linear relationship between the feature and decision spaces, which is not suitable for complex and large-scale data. To overcome the limitation, the FDC model is proposed as a novel embedded feature selection approach grounded in granular computing theory. By incorporating a fuzzy consistency metric, the FDC model enables nonlinear mapping from the feature space to the decision space, capturing the intricate relationship between features and decisions. FDC integrates a dual mechanism of fusion fuzzy information decision learning and classification reward, allowing it to simultaneously account for both the fuzzy and explicit aspects of classification. Experimental evaluations on 15 real-world datasets demonstrate that the FDC model effectively improves feature selection accuracy, suggesting its potential and applicability in practical settings.</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

Dual-driven embedded feature selection method based on fuzzy decision consistency and classification reward mechanism

  • 期刊论文

  • 发表日期：

    2025年3月15日

查看全文
  • Yang Huang,
  • Tingquan Deng,
  • Changzhong Wang,
  • Yang Zhang

摘要

Feature selection is vital in machine learning and data analysis, as it enhances model performance, reduces computational costs, and improves efficiency. However, existing embedded methods, particularly those based on regression, often assume a linear relationship between the feature and decision spaces, which is not suitable for complex and large-scale data. To overcome the limitation, the FDC model is proposed as a novel embedded feature selection approach grounded in granular computing theory. By incorporating a fuzzy consistency metric, the FDC model enables nonlinear mapping from the feature space to the decision space, capturing the intricate relationship between features and decisions. FDC integrates a dual mechanism of fusion fuzzy information decision learning and classification reward, allowing it to simultaneously account for both the fuzzy and explicit aspects of classification. Experimental evaluations on 15 real-world datasets demonstrate that the FDC model effectively improves feature selection accuracy, suggesting its potential and applicability in practical settings.