Scholarlink

<p>This study evaluates the performance of various machine learning based regression models, including Lasso, Ridge, Elastic Net, Non-Parametric, and Linear Regression, in predicting compressive strength (CS). The models were assessed using multiple performance metrics such as Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), R-squared (R²), and others. CS trends highlight the influence of curing time and the fly ash-to-waste plastic ratio. Early curing periods (3 and 7 days) show lower strength, which improves significantly over longer durations (28, 56, and 90 days) due to hydration and pozzolanic reactions. Non-Parametric model consistently outperformed the parametric models, achieving the lowest MAE (2.791 during training and 2.436 during testing) and the highest R² value (0.859 in training and 0.863 in testing), highlighting its superior predictive capability. Monotonicity analysis of the Non-Parametric model revealed that CS exhibits a strictly decreasing relationship with % Waste Plastic (R² = 0.911) and a strictly increasing relationship with % Fly Ash (R² = 1), while the relationship with Curing Days follows a non-monotonic trend (R² = 0.6837). Sensitivity analysis further demonstrated that % Plastic Waste has the highest impact on CS (sensitivity = 0.583), followed by Curing Days (0.414), whereas % Fly Ash has a minimal effect (0.002). These findings suggest that Non-Parametric models are more effective in capturing complex relationships in CS prediction, offering valuable insights for optimizing concrete compositions.</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

Prediction of compressive strength of concrete doped with waste plastic using machine learning-based advanced regularized regression models

  • 期刊论文

  • 发表日期：

    2025年2月14日

查看全文
  • Anish Kumar,
  • Sameer Sen,
  • Sanjeev Sinha,
  • Bimal Kumar,
  • Chaitanya Nidhi

摘要

This study evaluates the performance of various machine learning based regression models, including Lasso, Ridge, Elastic Net, Non-Parametric, and Linear Regression, in predicting compressive strength (CS). The models were assessed using multiple performance metrics such as Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), R-squared (R²), and others. CS trends highlight the influence of curing time and the fly ash-to-waste plastic ratio. Early curing periods (3 and 7 days) show lower strength, which improves significantly over longer durations (28, 56, and 90 days) due to hydration and pozzolanic reactions. Non-Parametric model consistently outperformed the parametric models, achieving the lowest MAE (2.791 during training and 2.436 during testing) and the highest R² value (0.859 in training and 0.863 in testing), highlighting its superior predictive capability. Monotonicity analysis of the Non-Parametric model revealed that CS exhibits a strictly decreasing relationship with % Waste Plastic (R² = 0.911) and a strictly increasing relationship with % Fly Ash (R² = 1), while the relationship with Curing Days follows a non-monotonic trend (R² = 0.6837). Sensitivity analysis further demonstrated that % Plastic Waste has the highest impact on CS (sensitivity = 0.583), followed by Curing Days (0.414), whereas % Fly Ash has a minimal effect (0.002). These findings suggest that Non-Parametric models are more effective in capturing complex relationships in CS prediction, offering valuable insights for optimizing concrete compositions.