Implementation of meta-ensembled algorithms via light gradient boosting model for predicting high-performance concrete main strength features

High-performance concrete compressive and tensile strengths are essential in terms of the assurance of structural performance and reliability. The research will describe the effective estimation of such properties through an artificial intelligence-based approach to overcome several limitations of experimental testing. For this purpose, a Light Gradient Boosting model has been developed and enhanced using four meta-heuristic optimization algorithms: Dandelion Optimization, Runge-Kutta Optimization, Seagull Optimization Algorithm, and Black Widow Optimization Algorithm. The LGRDSB was an ensemble model that combined the strengths of all four optimizers. Among them, the RUN optimizer with the LGRK model emerged as the best, giving R-squared values of 0.9928 and 0.9914 for CS and TS predictions, respectively. Thus, the LGRDSB model ensemble emerged as most robust and reliable to handle diverse datasets, securing R-squared values greater than 98% and less than 1% error rates. These results highlight the performance of the proposed models in predicting HPC properties and provide a realistic approach toward integrating AI techniques into performance evaluation for HPC.