Table 2.

Optimized hyperparameter and cross-validated MAE and R² for both data sets.

Model	Cross-validated training MAE (GPa)		Cross-validated test MAE (GPa)		Cross-validated training R2		Cross-validated test R2
	Refractory and non-refractory dataset	Refractory dataset	Refractory and non-refractory dataset	Refractory dataset	Refractory and non-refractory dataset	Refractory dataset	Refractory and non-refractory dataset	Refractory dataset
Gradient Boosting	0.42 ± 0.26	0.36 ± 0.16	10.37 ± 1.59	6.15 ± 1.19	0.99 ± 0.003	0.99 ± 0.007	0.71 ± 0.080	0.90 ± 0.036
XGBoost	0.33 ± 0.28	1.04 ± 0.48	10.32 ± 1.50	6.68 ± 1.22	0.99 ± 0.003	0.99 ± 0.008	0.70 ± 0.076	0.89 ± 0.038
RF	5.63 ± 0.59	5.54 ± 0.63	13.53 ± 1.50	9.00 ± 1.08	0.95 ± 0.009	0.96 ± 0.010	0.68 ± 0.076	0.89 ± 0.031
Ada Boost	12.79 ± 0.94	5.54 ± 0.84	18.02 ± 1.57	9.31 ± 1.53	0.86 ± 0.021	0.97 ± 0.011	0.62 ± 0.080	0.88 ± 0.051
SVM	14.78 ± 1.61	1.90 ± 0.57	17.83 ± 1.99	6.41 ± 1.39	0.64 ± 0.060	0.97 ± 0.013	0.54 ± 0.074	0.87 ± 0.053
Lasso regression	19.29 ± 1.44	17.53 ± 1.14	21.09 ± 1.64	18.16 ± 1.41	0.60 ± 0.060	0.72 ± 0.049	0.51 ± 0.076	0.67 ± 0.172
Ridge regression	19.37 ± 1.40	33.18 ± 3.32	21.24 ± 1.95	33.34 ± 3.26	0.60 ± 0.057	0.075 ± 0.007	0.51 ± 0.082	0.018 ± 0.065
Gaussian process	33.52 ± 1.90	34.08 ± 3.28	33.81 ± 1.92	34.55 ± 3.32	4.95 E−6 ± 4.9 E−7	1.35 E−5 ± 2.2 E−6	0.04 ± 0.028	0.090 ± 0.067