A Predictive Model for Guillain–Barré Syndrome Based on Ensemble Methods

Table 3

Average results of ensemble methods across 30 runs in four GBS subtype classification.

Ensemble method	Average accuracy	Multiclass AUC	Sensitivity	Specificity	Kappa
Random Forest	0.9366	0.8390	0.8120	0.9544	0.8090
	0.0245	0.0803	0.0812	0.0178	0.0748
C5.0	0.9272	0.8398	0.8126	0.9476	0.7825
	0.0251	0.0789	0.0749	0.0191	0.0746
Boosting	0.9195	0.8099	0.7906	0.9422	0.7596
	0.0202	0.0578	0.0648	0.0158	0.0610
Random Subspace	0.9016	0.7871	0.6607	0.9251	0.6960
	0.0216	0.0592	0.0691	0.0169	0.0682
Bagging	0.8980	0.7895	0.6936	0.9251	0.6923
	0.0284	0.0484	0.0622	0.0206	0.0831