Study on Concrete Strength Prediction Model Based on Gradient Boosting Regressor

DOI: https://doi.org/10.62517/jes.202502308

Author(s)

Mingzhen Li

Affiliation(s)

China Nuclear Power Engineering Co., Ltd., Xudapu Project Department, Xingcheng, Liaoning, China

Abstract

Concrete strength is a core indicator for evaluating the quality of construction projects, and its accurate prediction holds great significance for engineering design and construction quality control. However, traditional prediction methods (e.g., empirical formulas and linear regression) struggle to accurately capture the nonlinear relationships between multiple influencing factors and concrete strength. This study proposes a concrete strength prediction model based on Gradient Boosting Regressor (GBR), systematically elaborating its mathematical principles and parameter design logic. Twelve key features—including cement dosage, fly ash content, water-binder ratio, and age—are utilized to construct the prediction model. Through training and validation on 131 sets of concrete mix proportion and strength data from an actual engineering project, the model demonstrates excellent predictive performance: its coefficient of determination (R²) reaches 0.9063, root mean squared error (RMSE) is 5.0697 MPa, and mean absolute error (MAE) is 4.0340 MPa. The results indicate that the established GBR model can effectively capture the nonlinear relationships between concrete components and strength, providing a scientific basis for concrete mix proportion design and a reference for similar nonlinear prediction problems.

Keywords

Concrete Strength Prediction; Machine Learning; Gradient Boosting Regressor; Mix Proportion Optimization

References

[1] Li, J., Zhang, L., and Wang, Q. Machine learning-based prediction of concrete strength: A review. Construction and Building Materials, 2023, 367:130123. [2] Wang, Y., Liu, X., and Chen, H. Comparative study of machine learning models for predicting compressive strength of recycled aggregate concrete. Journal of Cleaner Production, 2022, 357:131945. [3] Zhang, S., Li, M., and Zhang, J. Prediction of concrete strength using gradient boosting decision tree and SHAP value analysis. Structural Concrete, 2021, 22(5):2345-2360. [4] Liu, H., Zhang, Y., and Wu, C. A hybrid machine learning model for predicting the compressive strength of high-performance concrete. Automation in Construction, 2020, 117:103287. [5] Chen, J., Yu, R., and Li, T. Prediction of concrete strength using random forest algorithm optimized by particle swarm optimization. Materials Research Express, 2023, 10(4):045701. [6] Zhao, X., Liu, J., and Wang, Y. Compressive strength prediction of recycled aggregate concrete using artificial neural network: A comparative study. Journal of Materials in Civil Engineering, 2022, 34(8):04022184. [7] Peng, Y., Zhang, L., and Li, Q. Gradient boosting regression for predicting the compressive strength of concrete with supplementary cementitious materials. Materials and Structures, 2021, 54(3):109. [8] Paul, S. C., Singh, B., and Reddy, K. H. Prediction of compressive strength of concrete using gene expression programming. Advances in Cement Research, 2020, 32(6):245-258. [9] Li, C., Wang, Q., and Zhang, Y. BP neural network optimized by genetic algorithm for predicting the compressive strength of self-compacting concrete. Neural Computing and Applications, 2023, 35(15):10897-10910. [10]Zhang, H., Liu, Y., and Chen, W. Support vector machine for predicting the compressive strength of high-strength concrete: A comparative study. KSCE Journal of Civil Engineering, 2022, 26(2):765-776. [11]Wang, Z., Li, H., and Zhang, J. Comparative study of ensemble learning methods for predicting the compressive strength of concrete. Journal of Testing and Evaluation, 2021, 49(2):1123-1140. [12]Chen, L., Zhang, Y., and Liu, J. A convolutional neural network approach for predicting the compressive strength of concrete. Automation in Construction, 2023, 147:104604. [13]Liu, X. Y., Liu, F. Y., Wang, Z. X., et al. Prediction of sulfate resistance of recycled concrete using four ensemble learning methods. Construction and Building Materials, 2022, 314:125637.