Investigation of Support Vector Machines with Different Kernel Functions for Prediction of Compressive Strength of Concrete

In this study, our primary aim is to assess and compare the efficacy of Support Vector Machines (SVM) employing various kernel functions: linear (LIN), polynomial (POL), Radial Basis Function (RBF), and sigmoid (SIG) in predicting the compressive strength of concrete. We generated and validated different models, namely SVM-LIN, SVM-POL, SVM-RBF, and SVM-SIG. Utilizing a dataset comprising 236 samples from the Red River surface water treatment plant project in Hanoi, Vietnam, we partitioned the data into training (70%) and testing (30%) sets for model training and validation. Our analysis employed various validation metrics, including coefficient of correlation (R), Root Mean Squared Error (RMSE), and Mean Absolute Error (MAE), to assess and compare model performance. Results indicate that SVM-RBF (R = 0.847) outperforms the other models on testing data, followed by SVM-POL (R = 0.7182), SVM-LIN (R = 0.6679), and SVM-SIG (R = 0.0198), respectively. Consequently, our findings suggest that the RBF kernel function is most suitable for training SVM models to predict concrete compressive strength. Therefore, SVM-RBF emerges as a promising tool for the rapid and accurate estimation of concrete compressive strength. This study contributes novel insights by systematically evaluating these models using a comprehensive set of validation metrics, enhancing the robustness and applicability of predictive models in concrete technology.