Predicting the Compressive Strength of Concrete: a Multivariate Approach Utilizing Machine Learning Regressors
Concrete’s compressive strength is a crucial factor in the design of new concrete structures and the assessment of existing ones in the civil engineering discipline. Precise estimation of concrete’s compressive strength is crucial to the advancement and quality of construction. Traditionally, estimating concrete’s strength characteristics has required a great deal of laboratory testing. To expedite this process, researchers are increasingly using predictive models. The current study is to apply the machine learning regressor models to predict the compressive strength of concrete. The multivariate analysis comprises nine independent variables and compressive strength of concrete as a dependent variable. For training and assessing the models, a dataset with 1133 test results of compressive strength of concrete is collected and used. The methodology employs XGBoost, CatBoost, LightGBM, and ensemble/stacked regressor models to predict the compressive strength using nine input parameters. The ensemble/stacked regressor outperforms the other three regressor models, achieving a root mean. squared error (RMSE) of 3.87 MPa, a mean absolute error (MAE) of 2.81 MPa, and a coefficient of determination (R2) of 0.955. The results of the models were validated statistically with different pairs of models, and the ensemble/stacked model significantly outperforms the individual regressor models. Finally, a web-based prediction interface was developed.
Сведения
- Опубликовано
- 2026 · Т. 4, № 2, С. 27-35
- Язык
- Английский
- ISSN
- 2181-4473
- Издатель
- Fergana State Technical University
- Аффилиации
- NICMAR University of Construction Studies, Hyderabad, Telangana, India
MBA ACM, NICMAR University of Construction Studies, Hyderabad, Telangana, India