Table B3.

Comparisons of median C-indices of six machine learning algorithms across varying censoring rates (40% to 80%) under a nonlinear log hazard model with various proportions of missing data for each covariate (0%, 10%, 20%). The table displays median C-index values from 100 experiments, along with their interquartile ranges.

	Censoring Rate
	40%	60%	70%	80%
0% Missing
Cox Model with Linear Log Hazards	65.6 (2.0)	67.8 (3.0)	69.0 (2.9)	72.5 (3.1)
Survival Support Vector Machines	73.0 (1.7)	74.6 (1.4)	76.3 (1.8)	78.7 (2.4)
Survival Gradient Boosting	72.6 (1.3)	74.2 (1.6)	75.7 (2.1)	78.0 (2.3)
Random Survival Forests	71.5 (1.5)	73.3 (1.6)	75.2 (2.0)	77.5 (1.9)
Naive Ensemble Averaging	71.5 (1.5)	73.4 (1.6)	75.3 (2.0)	77.6 (1.9)
Weighted Ensemble Averaging	73.1 (1.1)	74.8 (1.4)	76.6 (1.6)	78.9 (1.6)
10% Missing
Cox Model with Linear Log Hazards	64.5 (1.8)	66.3 (3.3)	67.9 (3.5)	70.8 (2.8)
Survival Support Vector Machines	70.7 (1.9)	72.3 (2.2)	73.5 (1.7)	75.9 (3.0)
Survival Gradient Boosting	69.9 (2.1)	71.4 (2.0)	73.0 (2.8)	74.7 (2.7)
Random Survival Forests	69.6 (1.6)	71.3 (2.5)	72.8 (2.8)	75.1 (2.3)
Naive Ensemble Averaging	69.6 (1.6)	71.4 (2.5)	72.8 (2.9)	75.2 (2.4)
Weighted Ensemble Averaging	70.9 (1.4)	72.5 (2.1)	74.4 (2.4)	76.4 (2.5)
20% Missing
Cox Model with Linear Log Hazards	63.8 (2.0)	65.6 (3.2)	66.0 (3.4)	69.4 (4.0)
Survival Support Vector Machines	68.9 (2.1)	70.6 (2.4)	71.5 (2.3)	73.5 (3.9)
Survival Gradient Boosting	67.8 (1.9)	70.1 (2.9)	70.0 (3.3)	72.9 (3.9)
Random Survival Forests	68.1 (2.0)	69.7 (2.0)	70.9 (3.3)	73.0 (3.9)
Naive Ensemble Averaging	68.1 (2.0)	69.8 (2.0)	71.0 (3.3)	73.1 (4.0)
Weighted Ensemble Averaging	69.2 (1.5)	71.0 (2.2)	72.0 (2.9)	74.5 (3.5)