Eine Plattform für die Wissenschaft: Bauingenieurwesen, Architektur und Urbanistik
Eine Plattform für die Wissenschaft: Bauingenieurwesen, Architektur und Urbanistik
Compressive strength prediction of high-strength concrete using hybrid machine learning approaches by incorporating SHAP analysis
Concrete is the most extensively used construction material, and cement is its main component. Hybrid machine learning models attract researchers in building materials due to their high applications and prediction accuracy. Hybrid machine learning model interpretability is crucial to apply to the interest of field experts. Therefore, this research study proposes to predict the compressive strength of high-strength concrete (HSC) using supervised HML algorithms (XGBR-BR, SVR-RFR, and GBR-DTR), which is rarely seen in the literature. Additionally, SHAP—a novel black-box interpretation approach—was employed to input variables and elucidate the predictions. The comparison revealed that all selected hybrid ML models provide acceptable accuracy for compressive strength predictions. Moreover, the XGBoost-BR model exhibited superior performance R2= 0.99 for the training phase and R2= 0.96 for the testing phase. The average error ranges were found to be very closely approximate ± 5. The predictions of the XGBoost-BR model capture significant correlations between the input variables based on the interpretation of SHAP. On the other hand, SHAP offers consistent measurements of a feature’s significance and a variable’s influence on a prediction. It is interesting to note that the SHAP interpretations matched what is typically observed in the compressive behavior of concrete, verifying the causality of the hybrid ML predictions. Applying hybrid machine learning techniques to predict concrete’s compressive strength will benefit the area of civil engineering application.
Compressive strength prediction of high-strength concrete using hybrid machine learning approaches by incorporating SHAP analysis
Concrete is the most extensively used construction material, and cement is its main component. Hybrid machine learning models attract researchers in building materials due to their high applications and prediction accuracy. Hybrid machine learning model interpretability is crucial to apply to the interest of field experts. Therefore, this research study proposes to predict the compressive strength of high-strength concrete (HSC) using supervised HML algorithms (XGBR-BR, SVR-RFR, and GBR-DTR), which is rarely seen in the literature. Additionally, SHAP—a novel black-box interpretation approach—was employed to input variables and elucidate the predictions. The comparison revealed that all selected hybrid ML models provide acceptable accuracy for compressive strength predictions. Moreover, the XGBoost-BR model exhibited superior performance R2= 0.99 for the training phase and R2= 0.96 for the testing phase. The average error ranges were found to be very closely approximate ± 5. The predictions of the XGBoost-BR model capture significant correlations between the input variables based on the interpretation of SHAP. On the other hand, SHAP offers consistent measurements of a feature’s significance and a variable’s influence on a prediction. It is interesting to note that the SHAP interpretations matched what is typically observed in the compressive behavior of concrete, verifying the causality of the hybrid ML predictions. Applying hybrid machine learning techniques to predict concrete’s compressive strength will benefit the area of civil engineering application.
Compressive strength prediction of high-strength concrete using hybrid machine learning approaches by incorporating SHAP analysis
Asian J Civ Eng
Kashem, Abul (Autor:in) / Das, Pobithra (Autor:in)
Asian Journal of Civil Engineering ; 24 ; 3243-3263
01.12.2023
21 pages
Aufsatz (Zeitschrift)
Elektronische Ressource
Englisch
Environmentally Friendly Concrete Compressive Strength Prediction Using Hybrid Machine Learning
| DOAJ | 2022