Table 6. Comparing the performance of the proposed Ensemble Subspace kNN model against conventional machine learning models when using the GA+STAT feature set.

Results of 10-fold cross validation over 30 runs.

Proposed ensemble subspace kNN (EkNN) model (No. of learners (NL): 30; Subspace Dimension (SD): 16)
Parameters		AUC	ORP FPR	ORP TPR	ACC
NL: 30, SD:16	Mean	0.818	0.201	0.836	0.821
	Std.	0.021	0.027	0.021	0.020
Simple kNN model (Distance: Euclidean)
k		AUC	ORP FPR	ORP TPR	ACC	Acc. Diff. (EkNN vs. kNN)
2	Mean	0.768	0.241	0.730	0.751	+0.070
	Std.	0.119	0.160	0.393	0.128	−0.108
5	Mean	0.778	0.300	0.833	0.783	+0.038
	Std.	0.107	0.265	0.103	0.103	−0.083
10	Mean	0.753	0.371	0.845	0.758	+0.063
	Std.	0.137	0.350	0.120	0.131	−0.111
Support Vector Machine models
Kernel		AUC	ORP FPR	ORP TPR	ACC	Acc. Diff. (EkNN vs. SVM)
Linear	Mean	0.782	0.342	0.860	0.784	+0.037
	Std.	0.126	0.352	0.110	0.120	−0.100
Gaussian	Mean.	0.808	0.353	0.876	0.799	+0.022
	Std.	0.112	0.416	0.107	0.111	−0.091
Naive Bayes model
Predictor distributions		AUC	ORP FPR	ORP TPR	ACC	Acc. Diff. (EkNN vs. Naïve Bayes)
Normal	Mean.	0.695	0.132	0.455	0.662	+0.159
	Std.	0.169	0.163	0.493	0.181	−0.161