Table 1. Best classification accuracies for diverse situations (two different feature selection approaches, four different kinship filtered sets, and three classifiers).

Levels of accuracy were calculated by 10-fold cross-validation and expressed as the means ± 10-fold variance. Bold represents greater accuracy than other classifiers for each kinship-based filtered subset.

		Approach 1		Approach 2
Kinship	Algorithm	# of Features	Mean ± Variance	# of Features	Mean ± Variance
	LogitBoost	83	0.652 ± 0.002	81	0.661 ± 0.004
≥ 0.00	KNN (IBk)	80	0.557 ± 0.006	90	0.549 ± 0.001
	SVM (SMO)	86	0.588 ± 0.004	87	0.578 ± 0.001
	LogitBoost	72	0.878 ± 0.002	85	0.868 ± 0.005
≥ 0.05	KNN (IBk)	88	0.720 ± 0.015	81	0.726 ± 0.010
	SVM (SMO)	88	0.784 ± 0.004	90	0.747 ± 0.009
	LogitBoost	47	0.950 ± 0.002	64	0.942 ± 0.003
≥ 0.10	KNN (IBk)	24	0.833 ± 0.013	28	0.850 ± 0.005
	SVM (SMO)	73	0.792 ± 0.009	50	0.790 ± 0.008
	LogitBoost	53	0.992 ± 0.001	4	0.992 ± 0.001
≥ 0.15	KNN (IBk)	82	0.983 ± 0.003	20	0.992 ± 0.001
	SVM (SMO)	73	0.967 ± 0.005	72	0.909 ± 0.007