Prediction and uncertainty quantification of compressive strength of high-strength concrete using optimized machine learning algorithms

Compressive strength is considered to be one of the most important mechanical properties of high-strength concrete (HSC). In this study, three machine learning models, ELM, PSO-ANN, and GS-SVR were employed to predict the compressive strength of HSC using 681 data records. The five ingredients and the compressive strength of HSC were regarded as input variable features and output target, respectively. Results indicated that the GS-SVR model showed the best performance in forecasting with the R of 0.992, MAPE of 0.016, RMSE of 1.241, MAE of 0.842, and RRMSE of 0.024, which could be an alternative candidate for strength prediction. Additionally, a sensitivity analysis was performed within the GS-SVR model to achieve an in-depth examination of the effect of each single input variable on the compressive strength. The sensitivity of input variables demonstrated that the water is the most sensitive parameter to the compressive strength of HSC, while the superplasticizer is less sensitive. The research of this paper can provide reference and guidance for concrete design and construction.