Scholarlink

<p>This work examines the ability of physics-inspired optimization techniques together with machine learning models by introducing eight versions of the Binary Henry Gas Solubility Optimization (B-HGSO) algorithm, based on gas solubility principles for feature selection in classification tasks. The Henry Gas Solubility Optimization (HGSO) Algorithm is a physical-based, global optimization approach initially designed for continuous :optimization problems. The B-HGSO algorithms incorporating S-shaped and V-shaped transfer functions to explore the feature space effectively, and using k-NN as a classifier, were evaluated on twelve UCI datasets, including Breast-cancer, Dermatology, and Lung-Cancer. These datasets provided a robust testbed for identifying the best-performing versions of the eight B-HGSO algorithms, which were subsequently compared against four well-established feature selection algorithms (B-GWO, B-PSO, B-MRFO, B-WOA) using accuracy, average number of features, and computational time as metrics. Statistical analysis using the Wilcoxon test demonstrated the significant performance of B-HGSO at a <InlineEquation ID="IEq1"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="11042_2025_20643_Article_IEq1.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="31" /> </InlineMediaObject> <EquationSource Format="TEX">\(95\%\)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>95</mn> <mo>%</mo> </mrow> </math></EquationSource> </InlineEquation> confidence level. The best-performing B-HGSO algorithm was then applied to a COVID-19 chest X-ray dataset as an effective tool for feature selection, utilizing the ability of GoogleNet and ResNet deep learning models for the feature extraction process. In this context, it was compared to B-PSO and B-GWO. This study highlights the effectiveness of B-HGSO in feature selection, particularly for medical image classification, offering a valuable tool for improving classification accuracy while reducing dimensionality.</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

An enhanced henry gas solubility optimization algorithm using transfer functions for feature selection problem

  • 期刊论文

  • 发表日期：

    2025年1月27日

查看全文
  • Soumaia Kahloul,
  • Djaafar Zouache

摘要

This work examines the ability of physics-inspired optimization techniques together with machine learning models by introducing eight versions of the Binary Henry Gas Solubility Optimization (B-HGSO) algorithm, based on gas solubility principles for feature selection in classification tasks. The Henry Gas Solubility Optimization (HGSO) Algorithm is a physical-based, global optimization approach initially designed for continuous :optimization problems. The B-HGSO algorithms incorporating S-shaped and V-shaped transfer functions to explore the feature space effectively, and using k-NN as a classifier, were evaluated on twelve UCI datasets, including Breast-cancer, Dermatology, and Lung-Cancer. These datasets provided a robust testbed for identifying the best-performing versions of the eight B-HGSO algorithms, which were subsequently compared against four well-established feature selection algorithms (B-GWO, B-PSO, B-MRFO, B-WOA) using accuracy, average number of features, and computational time as metrics. Statistical analysis using the Wilcoxon test demonstrated the significant performance of B-HGSO at a \(95\%\) 95 % confidence level. The best-performing B-HGSO algorithm was then applied to a COVID-19 chest X-ray dataset as an effective tool for feature selection, utilizing the ability of GoogleNet and ResNet deep learning models for the feature extraction process. In this context, it was compared to B-PSO and B-GWO. This study highlights the effectiveness of B-HGSO in feature selection, particularly for medical image classification, offering a valuable tool for improving classification accuracy while reducing dimensionality.