Scholarlink

<p>Feature selection is an optimization approach suitable for handling high-dimensional data with improved classification accuracy and fewer features. Metaheuristics are widely used for feature selection because of their simplicity and capability for global search. Grey Wolf Optimizer (GWO) is an emerging metaheuristic for various optimization problems. However, it suffers from a lack of balance between the wolves’ diversification and intensification capabilities when seeking prey and premature convergence. Thus, we proposed a hybrid Binary GWO (HLFC-BGWO) for well-balanced exploration and exploitation. Lévi flights are incorporated in the hunting phases of wolves to escape local optima, and chaos theory is used to maintain the equilibrium of local searching and finding the global solution by emulating a natural reflex action in the movement of wolves. Time-varying acceleration coefficients model the locations of search agents along with an exponential decay function for speedy convergence. Experimental results show that the overtaking percentage of HLFC-BGWO is 60.86% in benchmark functions compared to other algorithms and has better convergence speed than traditional GWO because of using the exponential decay function. Regarding IEEE CEC-C06 2019 functions, HLFC-BGWO demonstrates better results than other algorithms for all functions except CECF4, CECF8, and CECF9. The feature selection problem is addressed using various datasets, where HLFC-BGWO surpasses other metaheuristics in classification accuracy, feature selection, and average fitness. HLFC-BGWO also outperforms other metaheuristics in solving engineering design problems like Tension/Compression Spring, Welded Beam, Pressure Vessel, and Speed Reducer Design.</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

Fusion of Lévi flight and Chaos theory in binary Grey Wolf optimizer for feature selection and global optimization

  • 期刊论文

  • 发表日期：

    2026年6月15日

查看全文
  • Parijata Majumdar,
  • Sanjoy Mitra

摘要

Feature selection is an optimization approach suitable for handling high-dimensional data with improved classification accuracy and fewer features. Metaheuristics are widely used for feature selection because of their simplicity and capability for global search. Grey Wolf Optimizer (GWO) is an emerging metaheuristic for various optimization problems. However, it suffers from a lack of balance between the wolves’ diversification and intensification capabilities when seeking prey and premature convergence. Thus, we proposed a hybrid Binary GWO (HLFC-BGWO) for well-balanced exploration and exploitation. Lévi flights are incorporated in the hunting phases of wolves to escape local optima, and chaos theory is used to maintain the equilibrium of local searching and finding the global solution by emulating a natural reflex action in the movement of wolves. Time-varying acceleration coefficients model the locations of search agents along with an exponential decay function for speedy convergence. Experimental results show that the overtaking percentage of HLFC-BGWO is 60.86% in benchmark functions compared to other algorithms and has better convergence speed than traditional GWO because of using the exponential decay function. Regarding IEEE CEC-C06 2019 functions, HLFC-BGWO demonstrates better results than other algorithms for all functions except CECF4, CECF8, and CECF9. The feature selection problem is addressed using various datasets, where HLFC-BGWO surpasses other metaheuristics in classification accuracy, feature selection, and average fitness. HLFC-BGWO also outperforms other metaheuristics in solving engineering design problems like Tension/Compression Spring, Welded Beam, Pressure Vessel, and Speed Reducer Design.