AbstractSoil cation exchange capacity (CEC) strongly influences the chemical, physical, and biological properties of soil. As the direct measurement of the CEC is difficult, costly, and time-consuming, the indirect estimation of CEC from chemical and physical parameters has been considered as an alternative method by researchers. Accordingly, in this study, a new hybrid model using a support vector machine (SVM), coupling with particle swarm optimization (PSO), and integrated invasive weed optimization (IWO) algorithm is developed for estimating the soil CEC. The physical and chemical data (i.e., clay, organic matter (OM), and pH) from two field sites of Taybad and Semnan in Iran were used for validating the new proposed approach. The ability of the proposed model (SVM-PSOIWO) was compared with the individual model (SVM) and the hybrid model (SVM-PSO). The results of the SVM-PSOIWO model were also compared with those of existing studies. Different performance evaluation criteria such as RMSE, R2, MAE, RRMSE, and MAPE, Box plots, and scatter diagrams were used to test the ability of the proposed models for estimation of the CEC values. The results showed that the SVM-PSOIWO model with the RMSE (R2) of 0.229 Cmol + kg−1 (0.924) was better than those of the SVM and SVM-PSO models with the RMSE (R2) of 0.335 Cmol + kg−1 (0.843) and 0.279 Cmol + kg−1 (0.888), respectively. Furthermore, the ability of the SVM-PSOIWO model compared with existing studies, which used the genetic expression programming, artificial neural network, and multivariate adaptive regression splines models. The results indicated that the SVM-PSOIWO model estimates the CEC more accurately than existing studies.
Comparative analysis of support vector machine and artificial neural network models for soil cation exchange capacity prediction
