Table 3.

Averaged validation results across 10 random training-validation splits using Monte-Carlo cross-validation.

Condition	Binary accuracy (%)	AUC
1 (None–None)	64.1 ± 7.4	0.70 ± 0.08
2 (40%–10)	72.0 ± 18.5	0.80 ± 0.16
3 (40%–5)	76.3 ± 16.7	0.81 ± 0.16
4 (20%–10)	83.6 ± 13.6	0.90 ± 0.09
5 (20%–5)	67.9 ± 19.3	0.75 ± 0.20
6 (10%–10)	75.6 ± 11.7	0.81 ± 0.11
7 (10%–5)	59.9 ± 12.8	0.58 ± 0.23

Optimal training parameters are in bold.