Table 6. Support vector machine (SVM): Hyperparameters’ domain and the corresponding tuned values at the data sets under consideration.
The c, γ, ϵ, in respect, represents the regularisation term, kernel coefficient, and epsilon-tube (applicable only to linear SVM) value.
| method / data set | parameters | ||||
|---|---|---|---|---|---|
| kernels | degree | c | γ | ϵ | |
| SVM | {linear, poly, RBF, sigmoid} | {1, 2, 3} | [1, 4] | [-2.3, 0.7] | [-2.3, 0.7] |
| SVM at Demo | sigmoid | - | 0.333 | 0.504 | - |
| SVM at Fixation | poly | 3 | 1.640 | 0.563 | - |
| SVM at IA | rbf | 1 | 3.652 | 1.933 | - |
| SVM at Demo-Fixation | RBF | 2 | 3.929 | 1.941 | - |
| SVM at Demo-IA | poly | 3 | 3.950 | 1.987 | - |