An explanatory machine learning model for forecasting compressive strength of high-performance concrete

High-performance concrete (HPC) is one of the concrete types with high strength, good performance, and high durability, which has been considered in the structural industry. Testing and sampling this type of concrete to determine its mechanical properties is daunting and complex. In addition, human and environmental factors were significant in the preparation of samples, which was also time-consuming and energy-consuming. Artificial intelligence (AI) can be used to eliminate and reduce these factors. This article intends to use the machine learning (ML) method to forecast one of the HPC mechanical properties: compressive strength (CS). The experimental data set used from the published article includes 168 samples, of which 70% (118) of the sample belonged to training and 30% (50) to the testing phase. Least square support vector regression (LSSVR) is one of the ML models used for forecasting in this article. In addition, meta-heuristic algorithms have been utilized to obtain the target to improve the accuracy and reduce the error. Algorithms include Honey Badger algorithm (HBA), COOT optimization algorithm (COA), and generalized normal distribution optimization (GNDO). Combining the algorithms with the introduced model forms a hybrid format evaluated by metrics in the training and testing phases. By evaluating the hybrid models, it has been determined that they can forecast with high accuracy and are reliable. In general, the LSHB hybrid model obtained the highest R2 and the lowest error compared to other models.