Table 3.

Cross-validation (CV) results for dual graphs with k-NN (k = 1 to 5) and PAM

Training Set (2010)		Error Rate
Dual graphs (V)	Method	(10-fold CV) [%]
3-4	1-NN	8
(25 existing and 25 missing graphs)	2-NN	6
	3-NN	8
	4-NN	6
	5-NN	10
	PAM	34
3-5	1-NN	14
(43 existing and 43 missing graphs)	2-NN	12
	3-NN	15
	4-NN	13
	5-NN	14
	PAM	26
3-6	1-NN	13
(55 existing and 55 missing graphs)	2-NN	13
	3-NN	13
	4-NN	13
	5-NN	13
	PAM	26
3-7	1-NN	13
(61 existing and 61 missing graphs)	2-NN	13
	3-NN	13
	4-NN	13
	5-NN	12
	PAM	31
3-8	1-NN	13
(64 existing and 64 missing graphs)	2-NN	14
	3-NN	12
	4-NN	13
	5-NN	13
	PAM	33
3-9	1-NN	13
(68 existing and 68 missing graphs)	2-NN	13
	3-NN	13
	4-NN	13
	5-NN	12
	PAM	36