Incorporation of radial basis function with Gorilla Troops Optimization and Moth-Flame Optimization to predict the compressive strength of high-performance concrete

Current trends in modern research revolve around new technologies that can predict material properties without the expense of time, effort, and experimentation. Adapting machine learning methods to calculate various attributes of materials is receiving increasing attention. This study aims to forecast the 28-day compressive strength of high-performance concrete using both stand-alone and compound machine learning techniques. To this end, a stand-alone radial basis function and two ensemble optimizers, Gorilla Troops Optimization and Moth-Flame Optimization, have been applied. The R-2 (coefficient of determination), RMSE (root mean absolute error), MAE (mean absolute error), SI (scatter index), and NRMSE (normalized root mean squared error) cross-validation were used to validate the performance of each model. In addition, the input parameters' contribution to the outcomes' forecast is specified by using a sensitivity analysis. All techniques used have proven to show improved performance in predicting results. The RBF-MFO model was the most accurate, with an R-2 value of 0.996, compared to the RBF-GTO, with an R-2 value of 0.987. Moreover, in the RBF-MFO index, RMSE = 0.937, NRMSE = 0.0149, MAE = 0.1875, and SI = 0.0149. On the other hand, for the combined RBF-GTO model, RMSE = 1.9588, NRMSE = 0.0304, MAE = 0.8111, and SI = 0.0304. Based on the data obtained, it is clear that the combined RBF-MFO model has achieved better performance.