Prediction of high performance concrete (HPC) compressive strength is very important for determination of mix design optimization and structural reliability. An integrated deep learning framework, together with a Bayesian optimization for improving accuracy in prediction, is what this study presents. Finally, the optimized DNN yielded a coefficient of determination ( \(R^2\) ) of 0.932 and the lowest RMSE (4.104 MPa) among all the proposed regression models, and was superior to the traditional regression models, including linear regression (RMSE = 10.089 MPa) and kernel ridge regression (RMSE = 10.095 MPa). Hyperparameters were fine tuned using a Bayesian optimization reducing the RMSE from 10.985 MPa (unoptimized DNN) down to 4.104 MPa. SHAP-based feature importance analysis revealed that age and cement content were the most influential variables, reinforcing domain knowledge about cement hydration and strength development. In addition, a graphical user interface (GUI) was built to make practical implementation possible and thus allow real time compressive strength prediction using material proportions. In summary, it shows that a deep learning combined with hyperparameter optimization is able to greatly improve the predictive reliability and efficiency of the HPC strength, which facilitates the sustainable construction practices and efficient material utilization.
