Skip to main content
[Preprint]. 2023 Nov 21:arXiv:2311.11819v2. [Version 2]

Table III.

EVALUATION METRICS FOR ENSEMBLE METHOD PERMUTATIONS INCLUDING NUMBER OF BASE LEARNERS (TOP PART), COMPARTMENTALIZED VS. NON-COMPARTMENTALIZED BASE LEARNERS (MIDDLE PART), AND BASE LEARNERS OF VARYING ARCHITECTURES (BOTTOM PART)

Metric Model Aorta Cerebral Cardiac Average
Number of base learners RE ↓ Bagging-4 9.89% 28.19% 29.65% 22.57%
Bagging-8 9.71% 27.67% 29.20% 22.19%
Bagging-12 9.69% 27.22% 29.13% 22.01%
Stacking-4 10.22% 23.97% 32.24% 22.14%
Stacking-8 10.81% 25.18% 33.11% 23.03%
Stacking-12 10.77% 24.90% 35.81% 23.83%
RMSE ↓ Bagging-4 (1.46, 0.47, 0.61) (1.03, 0.86, 0.91) (2.23, 1.45, 1.99) (1.57, 0.93, 1.17)
Bagging-8 (1.41, 0.44, 0.60) (0.99, 0.85, 0.88) (2.15, 1.40, 1.86) (1.52, 0.90, 1.11)
Bagging-12 (1.41, 0.44, 0.59) (0.98, 0.83, 0.88) (2.13, 1.38, 1.85) (1.51, 0.88, 1.11)
Stacking-4 (1.28, 0.42, 0.48) (0.80, 0.76, 0.77) (2.09, 1.56, 1.84) (1.39, 0.91, 1.03)
Stacking-8 (1.39, 0.44, 0.59) (0.84, 0.78, 0.78) (2.16, 1.61, 1.97) (1.46, 0.94, 1.11)
Stacking-12 (1.40, 0.52, 0.57) (0.84, 0.76, 0.75) (2.10, 1.69, 1.96) (1.44, 0.99, 1.09)
k Bagging-4 (0.996, 0.972, 0.973) (0.908, 0.901, 0.897) (0.890, 0.872, 0.825) (0.931, 0.915, 0.898)
Bagging-8 (0.985, 0.968, 0.979) (0.900, 0.899, 0.899) (0.882, 0.868, 0.821) (0.922, 0.912, 0.900)
Bagging-12 (0.986, 0.964, 0.979) (0.899, 0.899, 0.900) (0.883, 0.867, 0.823) (0.923, 0.910, 0.901)
Stacking-4 (0.974, 0.959, 0.970) (0.918, 0.899, 0.924) (0.836, 0.827, 0.775) (0.909, 0.895, 0.890)
Stacking-8 (0.972, 0.958, 0.947) (0.925, 0.903, 0.906) (0.827, 0.789, 0.711) (0.908, 0.883, 0.855)
Stacking-12 (0.981, 0.954, 0.969 (0.937, 0.901, 0.947) (0.823, 0.768, 0.745) (0.914, 0.874, 0.887)
R2 Bagging-4 (0.983, 0.975, 0.981) (0.891, 0.910, 0.896) (0.879, 0.884, 0.790) (0.918, 0.923, 0.889)
Bagging-8 (0.983, 0.977, 0.981) (0.897, 0.911, 0.899) (0.885, 0.893, 0.815) (0.922, 0.927, 0.898)
Bagging-12 (0.983, 0.977, 0.981) (0.901, 0.915, 0.902) (0.887, 0.897, 0.817) (0.924, 0.930, 0.900)
Stacking-4 (0.984, 0.981, 0.987) (0.941, 0.925, 0.921) (0.881, 0.854, 0.804) (0.935, 0.920, 0.904)
Stacking-8 (0.981, 0.976, 0.982) (0.934, 0.913, 0.917) (0.861, 0.837, 0.782) (0.925, 0.909, 0.894)
Stacking-12 (0.980, 0.969, 0.981) (0.931, 0.919, 0.920) (0.861, 0.816, 0.780) (0.924, 0.901, 0.894)
Metric Model Aorta Cerebral Cardiac Average
Compart. vs. Non-compart. RE ↓ Bagging Comp-3 30.54% 30.56% 43.93% 35.01%
Bagging-3 9.86% 27.95% 29.77% 22.52%
Stacking Comp-3 11.23% 24.92% 35.56% 23.90%
Stacking-3 9.45% 24.14% 31.36% 21.65%
RMSE ↓ Bagging Comp-3 (11.66, 11.90, 10.28) (1.26, 1.77, 1.22) (12.81, 13.38, 11.87) (8.58, 8.81, 7.79)
Bagging-3 (1.38, 0.47, 0.62) (1.01, 0.86, 0.90) (2.25, 1.46, 1.96) (1.55, 0.93, 1.16)
Stacking Comp-3 (1.41, 0.45, 0.58) (0.82, 0.72, 0.77) (2.21, 1.65, 1.95) (1.48, 0.94, 1.10)
Stacking-3 (1.21, 0.39, 0.49) (0.79, 0.79, 0.77) (2.11, 1.50, 1.98) (1.37, 0.89, 1.08)
k Bagging Comp-3 (0.709, 0.638, 0.723) (0.851, 0.828, 0.836) (0.825, 0.779, 0.720) (0.795, 0.748, 0.760)
Bagging-3 (0.995, 0.965, 0.970) (0.916, 0.900, 0.894) (0.897, 0.874, 0.830) (0.936, 0.913, 0.898)
Stacking Comp-3 (0.982, 1.006, 1.005) (0.923, 0.920, 0.931) (0.845, 0.825, 0.785) (0.917, 0.917, 0.907)
Stacking-3 (0.971, 0.977, 0.987) (0.932, 0.902, 0.943) (0.857, 0.844, 0.770) (0.920, 0.908, 0.900)
R2 Bagging Comp-3 (0.954, 0.893, 0.927) (0.904, 0.905, 0.892) (0.807, 0.726, 0.679) (0.888, 0.841, 0.833)
Bagging-3 (0.984, 0.975, 0.981) (0.897, 0.911, 0.897) (0.880, 0.882, 0.799) (0.920, 0.923, 0.892)
Stacking Comp-3 (0.982, 0.976, 0.983) (0.935, 0.929, 0.918) (0.864, 0.825, 0.782) (0.927, 0.910, 0.894)
Stacking-3 (0.986, 0.982, 0.987) (0.940, 0.911, 0.920) (0.874, 0.865, 0.774) (0.933, 0.919, 0.894)
Metric Model Aorta Cerebral Cardiac Average
Arch. variation RE ↓ Bagging Blocks-3 10.35% 27.94% 31.28% 23.19%
Stacking Blocks-3 9.67% 23.77% 31.02% 21.48%
RMSE ↓ Bagging Blocks-3 (1.39, 0.51, 0.61) (1.13, 0.93, 0.90) (2.21, 1.49, 1.92) (1.58, 0.98, 1.14)
Stacking Blocks-3 (1.32, 0.38, 0.51) (0.80, 0.74, 0.76) (2.00, 1.47, 1.76) (1.37, 0.86, 1.01)
k Bagging Blocks-3 (0.999, 0.984, 0.977) (0.956, 0.947, 0.931) (0.910, 0.883, 0.812) (0.955, 0.938, 0.907)
Stacking Blocks-3 (0.980, 1.005, 0.968) (0.923, 0.909, 0.925) (0.866, 0.837, 0.814) (0.923, 0.917, 0.902)
R2 Bagging Blocks-3 (0.985, 0.977, 0.980) (0.901, 0.920, 0.903) (0.881, 0.890, 0.803) (0.922, 0.929, 0.895)
Stacking Blocks-3 (0.983, 0.982, 0.986) (0.936, 0.921, 0.917) (0.886, 0.874, 0.825) (0.935, 0.926, 0.909)

Each metric’s best value is highlighted in bold. Arrows indicate direction of improvement. RMSE given in cm/s. RMSE, k and R2 given by vx,vy,vz.