Hybrid Bayesian–grid search optimization for concrete strength prediction

The accurate prediction of concrete compressive strength is critical for ensuring structural reliability and optimizing material use. Recent advancements in machine learning (ML) offer promising alternatives in improving prediction accuracy and efficiency. However, many studies overlook the critical role of systematic hyperparameter optimization, which directly impacts model performance. This study addresses this gap by developing ML models enhanced through a novel Hybrid Bayesian-Grid Search (HBG) optimization technique, which integrates grid search with Bayesian optimization to efficiently identify optimal hyperparameters. HBG helps ensure comprehensive coverage while reducing evaluations by focusing on high-potential regions. Four ML algorithms, including Artificial Neural Networks (ANN), Support Vector Regression (SVR), K-Nearest Neighbors (KNN), and Multiple Linear Regression (MLR), were employed on a dataset comprising 1030 concrete samples. Feature importance is assessed using correlation analysis and the Boruta algorithm. Among those models, SVR achieved the highest prediction accuracy, with an R 2 of 0.939, outperforming ANN (R 2 = 0.924) and other models. The HBG optimization significantly enhanced model performance by identifying optimal hyperparameters. Variable importance revealed that cement, superplasticizer, and curing age are the most influential factors affecting concrete strength. Compared to existing studies, the optimized models developed in this research demonstrate superior accuracy and robustness. These findings highlight the effectiveness of ML in concrete strength prediction and the critical role of hyperparameter optimization in improving model reliability. The HBG technique offers a scalable and efficient approach for optimizing predictive models, contributing to cost-effective and time-efficient assessments in construction engineering.