Scholarlink

<p>This study introduces a framework combining metaheuristic optimization with Support Vector Machine (SVM) to predict California Bearing Ratio (CBR) in lateritic soils. Four optimization algorithms, namely, Additive Regression, Imperialist Competitive Algorithm, Artificial Bee Colony and Grey Wolf Optimization for SVM were trained on 144 soil samples characterized by eight geotechnical parameters. Results showed strong correlations between CBR and soil composition: negative with fines (<InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(r=-0.929\)</EquationSource> </InlineEquation>) and positive with gravel (<InlineEquation ID="IEq2"> <EquationSource Format="TEX">\(r=0.818\)</EquationSource> </InlineEquation>). The Artificial Bee Colony-Support Vector Machine (ABC-SVM) model demonstrated superior performance (training/testing <InlineEquation ID="IEq3"> <EquationSource Format="TEX">\(R^{2} = 0.9877/0.9531\)</EquationSource> </InlineEquation>), followed by Imperialist Competitive Algorithm-Support Vector Machine (ICA-SVM) and Grey Wolf Optimization-Support Vector Machine (GWO-SVM). ABC-SVM reduced testing RMSE by <InlineEquation ID="IEq4"> <EquationSource Format="TEX">\(20.3\%\)</EquationSource> </InlineEquation> versus conventional regression and exhibited greater stability with lower training–testing differentials (<InlineEquation ID="IEq5"> <EquationSource Format="TEX">\(\Delta R^{2} \approx 0.035\)</EquationSource> </InlineEquation> compared to <InlineEquation ID="IEq6"> <EquationSource Format="TEX">\(\Delta R^{2} &gt; 0.05\)</EquationSource> </InlineEquation>). The optimized models significantly improved prediction accuracy for lateritic soil CBR estimation.</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

Optimization-Enhanced Support Vector Machine Algorithms for California Bearing Ratio Prediction of Lateritic Soils: A Comparative Analysis

  • 期刊论文

  • 发表日期：

    2026年1月17日

查看全文
  • Pranshu Vardhan,
  • Suneet Kaur

摘要

This study introduces a framework combining metaheuristic optimization with Support Vector Machine (SVM) to predict California Bearing Ratio (CBR) in lateritic soils. Four optimization algorithms, namely, Additive Regression, Imperialist Competitive Algorithm, Artificial Bee Colony and Grey Wolf Optimization for SVM were trained on 144 soil samples characterized by eight geotechnical parameters. Results showed strong correlations between CBR and soil composition: negative with fines ( \(r=-0.929\) ) and positive with gravel ( \(r=0.818\) ). The Artificial Bee Colony-Support Vector Machine (ABC-SVM) model demonstrated superior performance (training/testing \(R^{2} = 0.9877/0.9531\) ), followed by Imperialist Competitive Algorithm-Support Vector Machine (ICA-SVM) and Grey Wolf Optimization-Support Vector Machine (GWO-SVM). ABC-SVM reduced testing RMSE by \(20.3\%\) versus conventional regression and exhibited greater stability with lower training–testing differentials ( \(\Delta R^{2} \approx 0.035\) compared to \(\Delta R^{2} > 0.05\) ). The optimized models significantly improved prediction accuracy for lateritic soil CBR estimation.