Predicting the capillary water absorption values of building stones by hybrid models

The determination of rock’s Capillary water absorption (CWA) requires extensive and complex empirical investigations, while forecasting algorithms could reduce the cost and time required. To achieve this goal, various rock records were collected from different rock types. For the forecasting model, two techniques named ANFIS and SVR were developed, with the model parameters determined using the Imperialist Competitive Algorithm (ICA). The proposed hybrid ICA-SVR and ICA-ANFIS models demonstrate exceptional performance in predicting CWA, with R^2 values exceeding 0.9851 and 0.9586 for the testing and training stages, respectively, indicating a strong correlation between predicted and measured CWA values. Considering all computed metrics, the SVR model optimized with ICA yielded better results than the ICA-ANFIS model in both testing and training stages. For example, the R^2 values for the ICA-SVR model were 0.9758 and 0.9747 for the testing and training datasets, respectively, with VAF values of 0.9889 and 0.9859 . The ICA-ANFIS model also produced acceptable results, though its performance was slightly weaker than the SVR model. Compared to a previous study, the proposed models show a significant improvement in efficiency, with the R^2 value increasing from 0.708 to 0.9889 . In summary, the enhanced ICA-SVR model can be considered a reliable and powerful tool for accurately determining the optimal values of the system’s key variables.