QuaBingo: A Prediction System for Protein Quaternary Structure Attributes Using Block Composition

Table 4

Performance comparison of classification algorithms in 10-fold cross-validation and self-consistency test.

Algorithms	Test method
	Cross-validation			Self-consistency
	CCI (%)	Kappa	-measure	CCI (%)	Kappa	-measure
Bayes
Bayes net	64.80	0.5017	0.608	65.02	0.5053	0.611
Naïve Bayes	39.91	0	0.228	39.91	0	0.228
Functions
LibSVM	67.40	0.5288	0.616	68.60	0.5464	0.632
Logistic	67.28	0.5285	0.615	67.57	0.5326	0.619
Multilayer perceptron	64.01	0.4893	0.598	69.97	0.5694	0.657
Lazy
IB1	51.91	0.3513	0.515	87.18	0.8218	0.869
IBk	58.45	0.4126	0.551	90.38	0.8682	0.902
KStar	62.43	0.463	0.581	88.10	0.8362	0.877
Meta
AdaBoostM1	59.80	0.3909	0.493	59.80	0.3909	0.493
Bagging	66.27	0.5147	0.605	69.88	0.5671	0.65
Rules
Conjunctive rule	59.80	0.3909	0.493	59.80	0.3909	0.493
Decision table	66.99	0.5189	0.601	67.22	0.5218	0.601
DTNB	67.12	0.5225	0.606	67.38	0.5268	0.61
Tree
J48	66.45	0.5161	0.607	69.81	0.5639	0.646
Random forest	58.91	0.4306	0.566	90.02	0.8651	0.899
Random tree	54.65	0.3817	0.537	90.38	0.8682	0.902

CCI is correctly classified instances.