Scholarlink

<p>The characteristics of penetration graded asphalt can be evaluated using various criteria, among which the penetration and softening point are considered critical. The rapid and accurate estimation of these parameters for graphene oxide (GO) modified asphalt can lead to significant time and cost savings. This study presents the first comprehensive application of Extreme Gradient Boosting (XGB) algorithm to predict these properties for GO modified asphalt, utilizing a diverse dataset (122 penetration, 130 softening point samples) from published studies. The developed XGB model, using 9 input parameters encompassing GO characteristics, mixing processes, and initial asphalt properties, demonstrated outstanding predictive accuracy (coefficient of determination R<sup>2</sup> of 0.995 on the testing data) and outperformed ten other benchmark machine learning algorithms. Furthermore, a Shapley Additive exPlanation (SHAP)-based analysis quantifies the feature importance, revealing that the base asphalt’s initial properties, aging type, and GO content are the most dominant predictive factors. This study delivers a validated, high-fidelity predictive tool that can significantly accelerate the material design process for GO-modified asphalt, providing an efficient and reliable alternative to extensive laboratory testing.</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 Effective Extreme Gradient Boosting Approach to Predict the Physical Properties of Graphene Oxide Modified Asphalt

  • 期刊论文

  • 发表日期：

    2025年10月1日

查看全文
  • Huong-Giang Thi Hoang,
  • Thuy-Anh Nguyen,
  • Hoang-Long Nguyen,
  • Hai-Bang Ly

摘要

The characteristics of penetration graded asphalt can be evaluated using various criteria, among which the penetration and softening point are considered critical. The rapid and accurate estimation of these parameters for graphene oxide (GO) modified asphalt can lead to significant time and cost savings. This study presents the first comprehensive application of Extreme Gradient Boosting (XGB) algorithm to predict these properties for GO modified asphalt, utilizing a diverse dataset (122 penetration, 130 softening point samples) from published studies. The developed XGB model, using 9 input parameters encompassing GO characteristics, mixing processes, and initial asphalt properties, demonstrated outstanding predictive accuracy (coefficient of determination R2 of 0.995 on the testing data) and outperformed ten other benchmark machine learning algorithms. Furthermore, a Shapley Additive exPlanation (SHAP)-based analysis quantifies the feature importance, revealing that the base asphalt’s initial properties, aging type, and GO content are the most dominant predictive factors. This study delivers a validated, high-fidelity predictive tool that can significantly accelerate the material design process for GO-modified asphalt, providing an efficient and reliable alternative to extensive laboratory testing.