Scholarlink

<p>This study presents a hybrid stacking ensemble framework integrated with explainable machine learning for predicting the load-carrying capacity of PVC tube-confined concrete columns under various eccentric loading conditions. A curated dataset encompassing geometric, material, and reinforcement parameters was used to train multiple base regressors-Support Vector Regression (SVR), Random Forest (RF), Gradient Boosting Regressor (GBR), XGBoost (XGB), and Artificial Neural Network (ANN). These were combined via a regularized linear meta-learner to optimize predictions. The proposed model achieved superior performance with an <InlineEquation ID="IEq1"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="42107_2025_1350_Article_IEq1.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="21" /> </InlineMediaObject> <EquationSource Format="TEX">\(R^2\)</EquationSource> <EquationSource Format="MATHML"><math> <msup> <mi>R</mi> <mn>2</mn> </msup> </math></EquationSource> </InlineEquation> of 0.9893, MAE of 7.71 kN, and RMSE of 10.56 kN, outperforming all individual base models. SHAP was interpreted to interpret the contribution of each input variable, identifying column height, eccentricity, and core diameter as the most significant features. Furthermore, a GUI was developed to make the model accessible for practical structural engineering applications. The results demonstrate that combining ensemble learning, explainable AI, and GUI deployment yields a highly accurate and interpretable system capable of aiding design and analysis of eccentrically loaded confined columns. </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

Hybrid machine learning modelling and feature interpretation of load-carrying capacity of PVC tube-confined concrete columns

  • 期刊论文

  • 发表日期：

    2025年5月8日

查看全文
  • Rupesh Kumar Tipu,
  • Vipin Kumar Verma

摘要

This study presents a hybrid stacking ensemble framework integrated with explainable machine learning for predicting the load-carrying capacity of PVC tube-confined concrete columns under various eccentric loading conditions. A curated dataset encompassing geometric, material, and reinforcement parameters was used to train multiple base regressors-Support Vector Regression (SVR), Random Forest (RF), Gradient Boosting Regressor (GBR), XGBoost (XGB), and Artificial Neural Network (ANN). These were combined via a regularized linear meta-learner to optimize predictions. The proposed model achieved superior performance with an \(R^2\) R 2 of 0.9893, MAE of 7.71 kN, and RMSE of 10.56 kN, outperforming all individual base models. SHAP was interpreted to interpret the contribution of each input variable, identifying column height, eccentricity, and core diameter as the most significant features. Furthermore, a GUI was developed to make the model accessible for practical structural engineering applications. The results demonstrate that combining ensemble learning, explainable AI, and GUI deployment yields a highly accurate and interpretable system capable of aiding design and analysis of eccentrically loaded confined columns.