Models | Hyperparameters |
---|---|
kNN | The number of predictors at each node = 1–10 |
RF | The number of predictors at each node = 105, the number of trees = 230 |
SVM (RBF) | The kernel width σ = 0.03125, the penalty parameter C = 2, and ε in the loss function = 0.05 |
RVM (Laplace) | The kernel width σ = 0.044 |
laGP | The initial values of lengthscale = 5, the initial values of nugget = 0.1 |
MPLE | The number of individual perceptrons = 18, the number of units in the hidden layer = 5–8 |
XGBoost | Step size shrinkage = 0.1, maximum depth of a tree = 7, the max number of iterations = 69 |