Research Article
Indirect Estimation of Swelling Pressure of Expansive Soil: GEP versus MEP Modelling
Table 6
Details of parameters settings along with various performance indices for RF, DT, SVM, and GB trees.
| RF | DT | SVM | GB-trees |
| No. of trees | 100 | Criterion | Least square | Kernel type | Dot | No. of trees and bins | 50 and 20 | Criterion | Least square | Maximal depth | 10 | C, epsilon ± | 0 | Maximal depth | 5 | Maximal depth | 10 | Prepruning | Yes | Converg epsilon | 0.001 | Min. rows | 10 | Prepruning | No | Min. gain | 0.01 | L Pos | 1 | Min. split improv | 1.00E-05 | Guess subset | Yes | Min. leaf size | 2 | L neg | 1 | Learning rate | 0.01 | RMSE | 60.612 ± 7.748 | RMSE | 74.694 ± 15.963 | RMSE | 90.382 ± 18.882 | RMSE | 76.828 ± 14.276 | Absolute error | 46.538 ± 5.798 | Absolute error | 57.52 ± 14.549 | Absolute error | 66.687 ± 11.900 | Absolute error | 62.154 ± 8.215 | Relative error | 53.09% ± 10.54% | Relative error | 59.21% ± 16.69% | Relative error | 65.67% ± 14.82% | Relative error | 73.03% ± 18.38% | RRSE | 0.650 ± 0.105 | RRSE | 0.812 ± 0.262 | RRSE | 0.949 ± 0.046 | RRSE | 0.809 ± 0.034 | Squared error | 3721.794 ± 966.481 | Squared error | 5783.066 ± 2326.161 | Squared error | 8454.147 ± 3641.390 | Squared error | 6065.551 ± 2325.052 | Correlation | 0.774 ± 0.083 | Correlation | 0.659 ± 0.165 | Correlation | 0.640 ± 0.104 | Correlation | 0.758 ± 0.071 | R2 | 0.604 ± 0.133 | R2 | 0.456 ± 0.213 | R2 | 0.419 ± 0.143 | R2 | 0.578 ± 0.107 | Spearman rho | 0.689 ± 0.096 | Spearman rho | 0.539 ± 0.177 | Spearman rho | 0.614 ± 0.112 | Spearman rho | 0.688 ± 0.070 | Kendall tau | 0.529 ± 0.086 | Kendall tau | 0.405 ± 0.150 | Kendall tau | 0.458 ± 0.098 | Kendall tau | 0.520 ± 0.071 |
|
|