Skip to main content
. 2021 Jul 22;121:108197. doi: 10.1016/j.patcog.2021.108197

Table 5.

Range of Hyperparameters (Hyp) for each model: HPC-XGB (our), Decision Tree (DT), Random Forest (RF), XGB with different losses functions (mean square error, tweedie and gamma), Linear Support Vector Machine (Li-SVM), Gaussian Support Vector Machine (G-SVM) and Lasso Support Vector Machine (La-SVM).

Model Hyp Range
HPC-XGB (our) learning rate # of estimators/iterations (EI) max depth (T) # of predictors to select l2 penalty (λ) {10−3,10−2,0.1,0.5}{5,15,25,50,75,100,125}{5,10,20,30,40,50,100,200} {5, 10, 15, 20, 25, 29} {10−3,10−2,0.1,0,1}
XGB [13], [14] learning rate # of estimators/iterations (EI) max depth (T) # of predictors to select l2 penalty (λ) {10−3,10−2,0.1,0.5}{5,15,25,50,75,100,125}{5,10,20,30,40,50,100,200} {5, 10, 15, 20, 25, 29} {10−3,10−2,0.1,0,1}
DT [10], [11] max depth {5,10,20,30,40,50,100,200}
RF [11], [12] # of DT # of predictors to select max depth {5,15,25,50,75,100,125}{5,10,15,20,25,29}{5,10,20,30,40,50,100,200}
Li-SVM [10], [11] Box Constraint {0.1,0.2,0.3,0.4,0.5,0.6,0.7,0.8,0.9,1}
G-SVM [10], [11] Box Constraint Kernel Scale {10−3,5·10−3,10−2,0.05,0.1,1,5,10,50,102}{10−3,5·10−3,10−2,0.05,0.1,1,5,10,50,102}
La-SVM [1] Lambda {0.1,0.2,0.3,0.4,0.5,0.6,0.7,0.8,0.9,1}
Ensemble La-SVM [15] Lambda {0.1,0.2,0.3,0.4,0.5,0.6,0.7,0.8,0.9,1}
l2,1 MTL [18] l2,1 penalty l1/l2 penalty {10−3,10−2,0.1,0.2,0.3,0.4,0.5,0.6,0.7,0.8,0.9,1}{10−3,10−2,0.1,0.2,0.3,0.4,0.5,0.6,0.7,0.8,0.9,1}